Servo Magazine - 2007 11

Page 84 7:56 PM 10/10/2007 Cover.qxd

Vol. 5 No. 11

SERVO MAGAZINE

TOUCH BIONICS • KILLER ROBOTS • HUMANOIDS • ANDROID ARMS

November 2007

4

74470 58285 0

1 1>

$7.00 CANADA

$5.50 U.S.

CoverInside.qxd

10/10/2007

11:34 AM

Page 2

We’re passive aggressive When it comes to passive products, we don’t pull any punches: we stock more major brands of passive components than any other major catalog distributor.* So whatever brands you need—

5

10

15

20

from AMP or AVX to Vishay or Wakefield— you’re more likely to find them all at Jameco. Check our “stats” below and see for yourself. It’s another Jameco advantage.

Jameco®

Digi-Key®

Mouser®

Newark®

AMP Astec Power Augat AVX Bourns Cherry Condor Power Corcom CTS Elco Connector Fox Electronics Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Power-One Teledyne Relays Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Corcom CTS Elco Connector Fox Electronics Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Power-One Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Condor Power Corcom Elco Connector Fox Electronics Int’l Rectifier Kemet Molex Panasonic Potter & Brumfield Teledyne Relays Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Corcom Elco Connector Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Vishay Wakefield

Free shipping on these and 83 other major brands. Call for details.

OTHER JAMECO ADVANTAGES:
More major brands of semis than any other catalog.
99% of catalog products ship the same day.
Lowest prices guaranteed, or we pay 10%.
Major brand names and generic equivalents

for even greater cost savings.

Order 24 hours a day, 7 days a week

www.Jameco.com Or call 800-831-4242 anytime ©Jameco Electronics. *According to the line cards on their web sites on August 28, 2007. Trademarks are the property of their respective owners.

Full Page.qxd

8/8/2007

7:54 PM

Page 3

:cignaZkZaXdcigdaaZg hj^iZYidhbVaagdWdi^Xh VcYXdcigdaVeea^XVi^dch

B^XgdbdjhZX]Vhh^h

GdWdcdkV"&

:VhnHiZe(%%%!VYkVcXZY hiZeeZgbdidgYg^kZVcY XdcigdahnhiZb

;^ghi"XaVhhXjhidbZghZgk^XZ### ¹Ndj ]VkZ egdk^YZY eda^iZ! YZiV^aZY hjeedgi l]Zc > ]VkZ cZZYZY^i!]VkZWZZcbdgZi]VcVXXdbbdYVi^c\l]Zc^iXdbZh idejgX]Vh^c\Vhh^hiVcXZVhlZaaVhegdk^Y^c\bZl^i]a^\]ic^c\ [VhiZbV^agZhedchZh!egdbeiYZa^kZgnVcY]VkZbZiVcY\gZVian ZmXZZYZYbnZmeZXiVi^dc[gdbVgZiV^aZg^cZkZgngZ\VgYº GdWZgiC:/ LLL#68I>K:"GD7DIH#8DB™I:A/ ))%&,+&'(.'+,

=D@JNDJG<"%)AMhiViZd[ i]ZVgiaVhZggVc\ZÃcYZg

=D@JNDE7H"%(?C >c[gVgZYgVc\ZÃcYZg[dg ^cYddgZck^gdcbZcih

6JI=DG>H:9=D@JND9:6A:G

GdWdiX]Vhh^h!eVgih VXXZhhdg^Zh

HZchdgBdjci^c\=VgYlVgZ

TOC Nov07.qxd

10/10/2007

4:58 PM

Page 4

PAGE 62

Columns

Departments

08

Robytes

06

Mind/Iron

10

GeerHead

07

Bio-Feedback

14

Ask Mr. Roboto

18

Events Calendar

56

Robotics Resources

20

New Products

31

Robotics Showcase

46

Robo-Links

Welcome to the (WowWee) Family

72

SERVO Webstore

Different Bits

82

Advertiser’s Index

by Jeff Eckert

Stimulating Robot Tidbits by David Geer

2007 FIRST Robotics Competition Winners by Pete Miles

Your Problems Solved Here by Gordon McComb

Learning Robotics From the World’s Robotics Labs

62 68

Twin Tweaks by Bryce and Evan Woolley

by Heather Dewey-Hagborg

Neural Networks for the PIC Microcontroller Part 3 — Hebbian Learning

76

Appetizer

78

Then and Now

4

by Bryan Bergeron

The Black Widow Contest Winner Humanoid Robots

SERVO 11.2007

by Tom Carroll

ENTER WITH CAUTION! 22 The Combat Zone

TOC Nov07.qxd

10/10/2007

5:20 PM

Page 5

11.2007 VOL. 5 NO. 11

Features & Projects 28

AUTOFLEX 2.0 by Brian Cieslak A new and improved autonomous programming tool for FIRST robots.

32

Killer Robots Are Our Friends

PAGE 32

by Brett Duesing A look inside the mechanics of combat robots.

36

Do You Want it Now? by Fred Eady Get instant gratification with the Firgelli PQ-CIB controller hooked up to their linear actuator.

42

Control a TOPO 1

PAGE 42

by Robert Doerr Breathe new life into an old robot.

47

Building an Android Arm

PAGE 47

by Mark Miller Part 2: Putting it all together.

51

GPS by Michael Simpson Part 2: Take a look at the Etek EB-85A, Copernicus, and Holux GPS modules.

PAGE 36 SERVO Magazine (ISSN 1546-0592/CDN Pub Agree #40702530) is published monthly for $24.95 per year by T & L Publications, Inc., 430 Princeland Court, Corona, CA 92879. PERIODICALS POSTAGE PAID AT CORONA, CA AND AT ADDITIONAL ENTRY MAILING OFFICES. POSTMASTER: Send address changes to SERVO Magazine, P.O. Box 15277, North Hollywood, CA 91615 or Station A, P.O. Box 54,Windsor ON N9A 6J5; [email protected]

SERVO 11.2007

5

Mind-Feed Nov07.qxd

10/10/2007

9:54 AM

Page 6

Published Monthly By T & L Publications, Inc. 430 Princeland Court Corona, CA 92879-1300 (951) 371-8497 FAX (951) 371-3052 Product Order Line 1-800-783-4624 www.servomagazine.com

Mind / Iron by Bryan Bergeron, Editor Œ In the commercial robotics world, all eyes are on the recent iRobot vs. Robot FX patent infringement lawsuits, in which iRobot is seeking to prevent Robot FX from selling any more Negotiator robots. While there are a number of facets to the case destined for the tabloids, one undisputed part of the story is that the suit comes on the heels of a competition between Robot FX and iRobot for a $280M contract with the US military. Robot FX won the contract. Whether iRobot — maker of the popular Packbot — gets another crack at the contract, the suit is important in that it marks an important milestone in the growth of the military robot industry. To follow my reasoning, consider the Gartner Hype Cycle, a popular model of technology-based products, first proposed by the Gartner Group (www.gartner.com) in 1995 (see Figure 1). According to the model, the first phase of a Hype Cycle is the “technology trigger,” marked by a significant breakthrough, public

VISABILITY

GARTNER HYPE CYCLE

Technology Trigger

Peak of Inflated Expectations

Trough of Disillusionment

Slope of Enlightenment

demonstration, product launch, and related events that generate press and industry interest. The next phase — the “Peak of Inflated Expectations” — is marked by over-enthusiasm and unrealistic expectations. In reality, there may be some successful applications of the technology, but there are more failures than winners. The only enterprises making money at this stage are conference organizers and magazine publishers. Following this over-hype and user/investor frustration from unmet expectations, technology-based products enter the “trough of disillusionment.” Because the press usually abandons the topic and the technology, this is the end for many products. Products that survive the trough of disillusionment – which may last months, years, or decades – are kept alive by companies that understand the technology’s applicability, risks, and benefits. The “slope of enlightenment” marks the time when there is practical, commercially-viable application of the technology – that is, FIGURE 1 some companies enjoy cash flow. Finally, the product and underlying technology reach the “plateau of productivity,” which is marked by the appearance of stable, accepted, second, and third generation products. Because it’s often difficult to directly track the few companies TIME Plateau of that are commercially Productivity successful during the Mind/Iron Continued

6

SERVO 11.2007

Subscriptions Inside US 1-877-525-2539 Outside US 1-818-487-4545 P.O. Box 15277 North Hollywood, CA 91615 PUBLISHER Larry Lemieux [email protected] ASSOCIATE PUBLISHER/ VP OF SALES/MARKETING Robin Lemieux [email protected] EDITOR Bryan Bergeron [email protected] CONTRIBUTING EDITORS Jeff Eckert Tom Carroll Gordon McComb David Geer Pete Miles R. Steven Rainwater Michael Simpson Kevin Berry Fred Eady Brett Duesing Brian Cieslak Mark Miller Robert Doerr James Baker Chad New Bryce & Evan Woolley Heather Dewey-Hagborg CIRCULATION DIRECTOR Tracy Kerley [email protected] MARKETING COORDINATOR WEBSTORE Brian Kirkpatrick [email protected] WEB CONTENT Michael Kaudze [email protected] PRODUCTION/GRAPHICS Shannon Lemieux Joe Keungmanivong ADMINISTRATIVE ASSISTANT Debbie Stauffacher

Copyright 2007 by T & L Publications, Inc. All Rights Reserved All advertising is subject to publisher’s approval. We are not responsible for mistakes, misprints, or typographical errors. SERVO Magazine assumes no responsibility for the availability or condition of advertised items or for the honesty of the advertiser.The publisher makes no claims for the legality of any item advertised in SERVO. This is the sole responsibility of the advertiser. Advertisers and their agencies agree to indemnify and protect the publisher from any and all claims, action, or expense arising from advertising placed in SERVO. Please send all editorial correspondence, UPS, overnight mail, and artwork to: 430 Princeland Court, Corona, CA 92879.

Mind-Feed Nov07.qxd

10/10/2007

9:55 AM

Page 7

“slope of enlightenment,” external events — such as lawsuits — serve as useful indicators. I’d like to propose this is the lawsuit point (shown in red in Figure 1) between the ‘trough of disillusionment” and “slope of enlightenment.” Historically, companies producing products aren’t bothered as long as they’re in an academic lab or smoldering in a company barely making a profit on the technology. However, as soon as the technology — and market — are mature enough to generate significant, sustainable revenue, then holders of patents (and their attorneys) take notice. The motivation for a suit may be strictly monetary. Some patent holders develop and hold on to a patent with no intent of developing a product. Instead, they hope that a technology will become viable before the term of their patent ends. A suit may be motivated by competition from a rival in the marketplace. In some cases, a suit is simply to establish the right of a company to compete in a given market. The iRobot–Robot FX suit suggests that the military robotics industry has survived the trough of disillusionment and is well on its way to the slope of enlightenment. There have been lawsuits in medical robotics, a sign that the robotics industry is making progress in this area, as well. How long before we see major lawsuits for home robots or assistive robots is unclear. However, when we do see lawsuits, it’ll be a sign that the field is maturing. Hopefully, the robotics companies involved in these suits will be financially fit enough to not only survive but thrive in the new economic environment. SV

Dear SERVO: Regarding the 09.2007 issue beginning on page 67, “Twin Tweaks — Robot vs.Wild” ... the problem stated was that the automotive steering vehicle had trouble making tight turns. The Wooleys solved part of the problem quite accurately with the Ackermann steering geometry, but you still have a solid rear axle (wheels, axle, and drive gears acting as a single unit). Thus, driving both rear wheels with relatively equal force when you try to turn, the front end gets pushed and you wind up going in a wider radius than the front wheels are set for. In the process of turning, the rear wheels want to slip because the outside wheel is traversing a larger arc than the inside wheel. If you’re going in a straight line, like drag racers do, a solid rear end is great. But if you want to make some turns, then you need a differential. And they almost had it — looking at the photo on the bottom of page 69 titled ‘Vex Differential.’ You need to cut the axle in two (a loose sleeve joining the two ends will allow independent motion and still keep the axles relatively concentric) and put a bevel gear on each axle end so that they mesh with the third bevel gear that’s attached to the differential carrier. This will allow continuous power to be applied to both rear wheels, irregardless of each wheel’s speed.You guys are doing great — hang in there.You’ll never know what you can do until you push your limits. — Phillip Potter continued on page 75

SERVO 11.2007

7

Robytes.qxd

10/9/2007

7:56 AM

Page 8

Robytes Autonomous Refueling Demonstrated

autonomously; the pilots shown in the photo were on board “for safety purposes.”

UAV for Farmers

The AARD system performs “better than a skilled pilot.” Photo courtesy of DARPA.

The Defense Advanced Research Projects Agency (DARPA, www. darpa.gov) has added to its bag of aeronautical tricks with the Autonomous Airborne Refueling Demonstration (AARD) program, through which it has demonstrated the first-ever robotic system to refuel airplanes in flight. In a recent series of tests, the AARD was fitted to a NASA-owned F/A-18 Hornet fighter and operated out of California’s Edwards Air Force Base. Using inertial, GPS, and video measurements — along with some special guidance and control techniques — the AARD managed to poke a refueling probe into a 32-inch basket while traveling 250 mph at 18,000 ft above the Tehachapi Mountains. Some tests were conducted in straight-and-level flight, under a range of turbulence conditions that involved as much as five feet of side-to-side movement of the drogue (the small windsock at the end of the refueling hose). In its most successful configuration, the AARD hit the target in 18 out of 18 attempts. It also managed to make the connection when the 707-300 tanker and F/A-18 were executing a turn, which is not usually attempted with a human pilot. In the tests, the fighter was operating

8

SERVO 11.2007

Most of the glory in the UAV arena goes to exotic military and security aircraft, but a fleet of miniature planes may soon create a buzz over the fields and forests of the heartland, providing surveillance for farming, environmental monitoring, and forestry. MicroPilot, Inc. (www.micro pilot.com), based in Stony Mountain, Manitoba, offers a range of UAVs, autopilots, and software products, including the MP-Vision airplane. Earlier this year, MicroPilot’s Crop Cam division (www.cropcam.com) introduced a version that has been configured specifically for agricultural operations. The CropCam AUV is a GPSguided craft that covers a preprogrammed flight pattern over a quarter section (160 acres) and takes digital photos along the way. With an overall length of four feet and a wingspan of eight feet, the six-pound plane can climb to 2,200 feet and complete a survey in about

Image taken by a CropCam AUV. Photo courtesy of Cropcam, Inc.

by Jeff Eckert 20 minutes. Guidance is provided by a Trimble GPS unit, and you can choose among three Pentax camera models to get up to eight megapixel resolution for stills and 640 x 480, 30 fps, in video mode. Power is provided by a 0.15 cu in engine that draws from a six-oz tank, but it appears that you can also get one that is driven by an Axi brushless motor and lithium polymer batteries. Rumor has it that it will run you about $7,000.

Bionic Hand Now Available

The i-LIMB Hand looks and acts like the real deal. Photo courtesy of Touch Bionics.

The Touch Bionics’ (www.touch bionics.com) i-LIMB Hand, formally introduced in July at the 12th World Congress of the International Society for Prosthetics and Orthoticsin Vancouver, Canada, looks like a great innovation for patients who are missing a hand through accidents, acts of war, or birth defects. Designed to look and operate like the real thing, it is said to be the world’s first commercially available prosthetic device with five individually powered digits. The device operates on an intuitive control system that uses a traditional myoelectric signal input to open and close its fingers. Myoelectric controls

Robytes.qxd

10/9/2007

7:57 AM

Page 9

Robytes use electrical signals generated by muscles in the remaining portion of a patient’s limb, with the signal being picked up by skin-mounted electrodes. Not shown in the photo is the available “cosmesis” covering, which makes it appear more lifelike in use. The device is already being fitted to patients in many clinics in the US and Europe.

Interactive Boybot

Build Your Own ROV

The ROV-in-a-Box kit comes more or less complete. Photo courtesy of !nventivity LLC.

It’s not pretty, but at least it’s pretty cheap. Designed for ages 12 and up, the ROV-in-a-Box kit from !nventivity (www.nventivity.com) sells for $249.95 and includes all of the required parts (frame, motors, light, camera, tether, controller, and battery), plus an instruction manual. It also comes with propellers, switches, connectors, “buoyancy devices” (presumably the chunks of plastic foam shown in the photo), and pretty much everything else. All you have to provide is PVC cement, tools, and a video monitor. According to the vendor, independent left and right props give it good controllability and zero-radius turning, and the light is bright enough to allow night missions. See the company’s website for a six minute video.

Zeno — a 17-inch mechanical boy — walks, talks, and interacts on a personal level. Photo courtesy of Hanson Robotics.

He looks quite a bit like the Japanese comic book character Astro Boy, but the new Zeno bot

from Hanson Robotics (www.hanson robotics.com) is actually named after the inventor’s son. Zeno’s main claim to fame is how well he imitates human facial expressions, but he also walks, talks, and can learn to recognize individual human beings (using a camera located behind one of his eyes) and address them by name. Like other Hanson creations (recall the familiar talking Einstein bot), Zeno is based on AI capabilities that help him learn and interact with his environment, a complex range (62, to be precise) of facial and neck expressions, his somewhat weird Frubber™ polymer skin, and the ability to develop a unique personality. According to Hanson, Zeno and his pals can be used in education, psychiatry, military training, and character development for animation. Some people find him adorable, and others have described him as “creepy,” so you’ll have to judge for yourself. Zeno is still a prototype, but the plan is to have a commercial version on the market in two years for about $300. SV

SERVO 11.2007

9

Geerhead.qxd

10/9/2007

7:53 AM

Page 10

by David Geer

Contact the author at [email protected]

2007 FIRST Robotics Competition Winners The FIRST (For Inspiration and Recognition of Science and Technology) Robotics Competition pits high school robotics teams against each other (and themselves!) with a different robot kit and task each season. n 2007, the Worcester Polytechnic Institute (WPI) supported Massachusetts Academy of Mathematics and Sciences at WPI (or MASS Academy, Team 190) won the FIRST Robotics World Championship in the Georgia Dome in Atlanta on April 14. Team 190 designed and constructed the winning robot — Goat-Dactyl — early in the season. Goat-Dactyl is a wheel-locomotive robot with sensors for autonomous control and R/C for remote. Team 190 designed the robot to accomplish specific, competition-related tasks as part of the FIRST 2007 competition. The robot completes the tasks as part of a game in competition and collaboration with other teams’ robots. This year’s competition game —

I

Students with Goat-Dactyl competition robot and control console queuing up before a match. The driver is thinking about strategy. Dan Jones, robot operator, is in the foreground and Colin Rody, driver, is in the background.

called “Rack ‘N’ Roll” — tested the students’ and their robots’ ability to (1) hang inflated colored tubes on pegs, configured in rows and columns, on a 10-foot-high center “rack” structure; (2) program a robotic vision system to navigate the robot; and (3) “lift” other robots more than 12 inches off the floor, according to Brad Miller, a Team 190 member. The leaders of the competition formed the aforementioned rack structure out of eight columns with three pegs each on which robot teams could place their tubes. “Every other column had a green light. The teams calibrated their robots’ cameras to track the light. Six robots took the field during a match. Officials assigned the robots Goat-Dactyl, mouth wide open, just before completing the lift of alliance partners. Dan Jones, operator, operating the controls in the background.

to either the blue or red alliance for competition. The teams earned points by hanging their alliance-colored tubes on one or more of the rack pegs,” says Miller. According to Miller, each hung tube was worth two points unless it was contiguous (either vertically or horizontally) with another hanging tube of your alliance color. “The total point count in this case was equal to two raised to the power corresponding to the length of the matched tube row or column (e.g., one tube = two points, two tubes = four points, three tubes = eight points ... a full circle of eight tubes = 256 points!),” Miller explains. Team 190 made the Goat-Dactyl robot from a kit that every team had to adhere to. The kit includes parts for the robot’s pneumatic and electrical systems, as well as a choice of motors. The robot itself consists of four CIM FR801-001 motors, which drive the robot. The large, broad metallic gripper that is the primary capability of the robot opens and closes with the aid of an RS-540 gear motor (Banebots). Two Globe 409A587 motors actuate the robot’s ramps. The team machined both the Photos are courtesy of Brad Miller, Team 190 member.

10

SERVO 11.2007

Geerhead.qxd

10/9/2007

7:53 AM

Page 11

GEERHEAD chassis and gripper elevator from scratch. They used 6061 aluminum C-channel and Lexan materials. They cut the lifting ramps via laser out of 5052 aluminum sheet metal. They used sheet metal in gauges ranging from .02” to .065” in thickness. Team 190 folded the aluminum ramps over, dimpled them with holes, and then riveted everything together. The gripper top is Lexan; the gripper bottom is fiberglass. The pneumatics included a Thompson compressor, accumulator tanks from Clippard, solenoid valves from SMC and Parker, and Bimba actuators. The actuator specs included three 1.5” bore by 3” stroke cylinders per side of the robot, which presented sufficient force to lift competing robots off the floor by more than a foot. The robot also featured a .75” bore by 8” stroke cylinder for grabbing onto and lifting the large inflated rings. Both of these maneuvers were useful for competition scoring.

Computer Controlled Each team is constrained to a kit that includes two PIC 18F8722 microprocessors. One is the slave and one is the master processor. The master processor controls the motor and communications and interfaces with the human operator. The slave contains all the original programming from the team’s coders. The robot passes some data between the master processor and the slave to process the actuators’ values. Team 190 coded the robot’s program in the C language. The coders used both the Microchip tools that come with the kit, and the Eclipse IDE. While teams in the FIRST competition can stick with Microchip’s tools that come with the kit, they are free to use other programming tools. “Our students use Eclipse as the development environment (IDE) for their in-class projects and are very familiar with it. So, we adapted some

Team 190 hanging a tube on the middle section of the rack.

work developed by other teams to make a development environment that suited us. Eclipse has a huge number of collaborators so even though it’s free, it is much higher quality than many of the commercial products,” says Miller. Team 190 also uses a software library named WPILib, which is a development framework that supports the standard FIRST devices like speed controls, the CMU camera, and gyros, for example.

Command and Control Team 190, as other teams, built a custom control system for interfacing with the competition-specified operator interface. That interface is the controls that enable the drivers and the robot to “talk” to each other. The FIRST supplied controller connects with joysticks, switches, potentiometers, and other control hardware. The controller transmits Team 190 members Dan Jones, operator, (back) and Paul Ventimiglia, mechanical lead (front) making last minute repairs on the robot between matches.

Team 190 putting a tube on the rack despite blocking attempts by a robot from the red alliance.

“control positions” to and from the robot. This enables the robot’s driver to manipulate the robot in competition while designing a unique set of controls for their purposes. Team 190 used two joystick controls for driving, and a separate control box of “arcade buttons” and switches to control the tube manipulator and robot-lift functions, according to Miller. Two operators handle the robots, one controlling the drive and the other controlling the manipulator and peripheral functions. The robot has many sensors, which help automate tasks such as lowering the robot’s lifts and raising the tube manipulator to

SELECTING ALLIES During selection for team allies before competition, Team 190 chose teams with compatible designs and tactics. “Through our excellent scouting and “intelligence” program, we were able to pick teams that we knew would make our alliance strong. Little did we realize that they would also make us look good, as well,” says Brad Miller, a Team 190 member. From among all the possibilities, Team 190 ended up collaborating with teams that all had maroon team shirt colors similar to their own. “Denying that matching team shirts was one of our selection criteria, we nonetheless took this as a good sign and have since celebrated this occurrence by producing “Don’t Mess With Maroon” championship shirts,” says Miller.

SERVO 11.2007

11

Geerhead.qxd

10/9/2007

7:54 AM

Page 12

GEERHEAD RESOURCES Worcester Polytechnic Institute www.wpi.edu/index.html WPI FIRST Robotics Resource Center http://first.wpi.edu WPI Robotics Engineering Major www.wpi.edu/Academics/Majors/RBE WPI Winning Robotics Team 190 http://users.wpi.edu/~first

Goat-Dactyl using its tube gripper to lift a tube during competition.

pre-specified heights. The goals for the robot were to endow it will the best abilities to win the game set before all the robots

while staying within competition limits. The major constraints for the robots include weight — 120 lbs and under — total size, the ability to recognize

The robot’s tube gripper is mounted on an extension mechanism (an elevator) that it uses to get it to the right height after it grabs the tube and sets it to the proper angle.

Team 190 between finals matches, on the field resetting the robot to play again while being overlooked by head ref, Aidan Browne.

Precision CNC Machining When you’re serious about hardware, you need serious tools. Whether milling 0.020” traces on prototype PCBs or cutting ½” steel battle armor, this CNC mill can do it all. Weighing in at more than 1100 lbs, the PCNC can deliver the hardware end of your combined hardware & software projects.

Tormach PCNC 1100 Features: Q

Table size 34" x 9.5"

Q

R8 Spindle 1.5 hp variable speed to 4500 RPM

Q

Computer controlled spindle speed and direction

Q

Precision ground ballscrews

Q

Digitizing and tool sensing support

Q

4th axis and high speed spindle options

3 Axis Mill

$6800 plus shipping

Mill includes Control, CAD and CAM software. Optional stand, coolant system, computer and accessories are extra.

Product information and online ordering at www.tormach.com

12

SERVO 11.2007

other robots, compete better than other robots, and to stay within cost limitations.

The Answers In response to these goals and limitations, Team 190 worked on lowering their bot’s weight while attaining its overall goals. “Our original robot-lift design was double the acceptable weight. We went through many different design approaches, including using aluminum honeycomb surfaces or making our own foam-core sheets, before finally settling on a unique sheet-metal box structure which was dimpled for improved strength. This same approach was used in nearly all aspects of the design,” says Miller. The tube gripper lays in front of the robot to grab tubes from the ground and catapult them high in the air to rack them up for scoring. (The tube gripper is in the front of the robot that can grab tubes from the ground and lift them to any height on the rack for scoring.) “Two of the unique features of the gripper are: It can grab tubes on the fly, without requiring the robot to come to a stop to pick them up; and second, it closes and lifts using a mechanism driven by a single pneumatic actuator. Usually two motions like this would require two actuators, but due to some clever design, only one is needed,” Miller explains. The team mounted the tube gripper on an extension mechanism (elevator) to get it to the right height after it grabs the tube and sets it to the proper angle, Miller further explains. The gripper is empowered by a single air cylinder that both closes the robot’s claw and raises the tubes up to a 55 degree angle in a single motion. “In addition,” says Miller, “the top digit of the claw is a four bar articulated linkage that curls around the tube, giving us maximum wrap, while allowing it to fit within our starting box.” SV

Full Page.qxd

10/10/2007

11:40 AM

Page 13

SERVO 11.2007

13

MrRoboto.qxd

10/9/2007

10:34 AM

Page 14

Our resident expert on all things robotic is merely an Email away.

[email protected]

Tap into the sum of all human knowledge and get your questions answered here! From software algorithms to material selection, Mr. Roboto strives to meet you where you are — and what more would you expect from a complex service droid?

by

Pete Miles

Q

. It is my understanding that the HSR-8498HB servos that are used in the Robonova humanoid robot have position feedback capabilities, so I bought a couple of them from Tower Hobbies. I have been trying for several days now to figure out how to get position data from these servos. From what I have seen on the Internet, all I have to do is send the servo a 50 microsecond pulse, and it will return a position signal that is similar to the regular pulse width to move the servo. I am missing

something here. Can you help me? — Pete Senganni

A

. The key to doing this is to use a pullup resistor on the signal line. This is required for bi-directional communication since the signal line is an open collector. Figure 1 shows a simple schematic for connecting an HSR-8498HB servo to a BASIC Stamp. Here, I used a 1K ohm resistor as a pullup resistor on the signal line. Without the resistor, you will not receive a signal back from the servo. Figure 2 shows a

sketch of the PWM (Pulse Width Modulation) control signal timing that is required for this servo to return its current position. To obtain the current position of the servo, you need to send a 50 µs pulse to the servo then wait for a minimum of 2 ms before measuring the width of the return pulse. The critical element required to measure the pulse width is to make sure that the servo signal line that is connected to the microcontroller is changed from an output signal line to an input signal line immediately prior to measuring

Figure 1. Connecting an HSR-8498HB servo to a BASIC Stamp for positional bi-directional control. +5V

SOUT

VDD VIN

SIN

RES

VSS

P1 P2 P3 P4

BASIC STAMP 2 FAMILY

P0

P5

14

P15

SIGNAL

P14

4.8-6.0V

P13

GND

SERVO

P12 P11 P10

P6 P7

+4.8 - 6.0V SERVO POWER 1 KΩ

ATN

P9 VSS

SERVO 11.2007

P8

HSR-8498HB SERVO AND CABLE

MrRoboto.qxd

10/9/2007

10:34 AM

Page 15

the pulse width. Or there will be no return signal from the servo (that you can see). The following BASIC Stamp sample code is all that is needed to measure the position feedback from the HSR-8498HB servo. This code sample assumes that the signal line is connected to P15 on the BASIC Stamp. The Delay_50us constant of 63 is the conversion factor for a 50 µs time delay on a BASIC Stamp 2px. This delay constant will be different for different Stamps. The “tmp*8/10” is a conversion factor to convert the PULSIN value back to microseconds on a BASIC Stamp 2px (again this conversion factor will be different for different BASIC Stamps). Tmp Position Servo_Pin Delay_50us

VAR VAR CON CON

WORD WORD 15 63

READ_PWM: LOW Servo_Pin PULSOUT Servo_Pin, Delay_50us PULSIN Servo_Pin, 1, tmp Position = tmp*8/10 DEBUG CRSRXY, 0,0,DEC5 POS PAUSE 5 GOTO READ_PWM

V HI

GND

50µs

500 - 2500 µs 2µs (MINIMUM)

OUTPUT FROM MICROCONTROLLER TO SERVO

OUTPUT FROM SERVO TO MICROCONTROLLER

Figure 2. PWM signal for position feedback for the HSR-8498HB servo. the right), the position will change on the debug window. Notice what I said here — when the servo horn is manually turned. Whenever the position is being read, the power to the internal motor of the servo is turned off, which allows the servo horn to be easily rotated by hand. This is actually a good feature to have when you are posing a humanoid robot for teaching new body positions. The drawback to this, however, is that the motor loses power for a moment when the position is being read. This may cause some servo jittering in some closed loop position control applications. The HSR-8498HB servo does require the position command pulse to be updated every 20 ms, like regular analog servos. If you continually read the servo’s position once between each position update cycle (i.e., once every 20 ms), the servo will jitter, and

Because of the pullup resistor, the logic state of the Servo_Pin will be high when it is not driven. Since the PULSOUT command on a Stamp toggles the current state of the output pin, it is manually set LOW prior to sending the 50 µs pulse to the servo, so the servo will recognize the 50 µs Operating Voltage 4.8V-6.0V positive pulse. On a No-load Speed @ 4.8V 0.26 sec/60° BASIC Stamp, the No-load Speed @ 6.0V 0.20 sec/60° PULSIN command autoStall Torque (4.8V) 84 oz/in (5.2 kg-cm) matically changes the pin’s state to an input Stall Torque (6.0V) 103 oz/in (7.4 kg-cm) state, so nothing speIdle Current Draw 8 mA cial needs to be done to No-load Current Draw, 200 mA measure the incoming Running pulse from the servo. Stall Current Draw 1200 mA This program will Pulse Range (~180°) 600-2350 µs continually read the Centering Pulse 1500 µs position of the servo and display the results x .78” x 1.85” Dimensions 1.57” (40 x 20 x 47 mm) on a debug window. When the servo horn is Weight 1.92 oz (54.7 g) manually rotated back Table 1. HSR-8498HB specifications. and forth (from left to

have about half the normal output torque. It is best to read the servo torque once every several position updates, such as once every five update cycles (or once every 100 ms). This will begin to minimize the amount of servo jitter and torque loss due to reading the servo position. For those that are not familiar with the HSR-8498HB servos, these servos are specifically designed for robotic applications. They look quite a bit different from standard R/C servos. Figure 3 shows a photo of two of them. One of the nice features of these servos is that they can be reconfigured for different applications, which includes a “bearing” joint at the bottom of the servo case, so that the main servo horn isn’t supporting the entire weight of the robot when the servo is used as a joint. Figure 3 shows one of these servos

Figure 3. HSR-9498HB servo configurations; traditional servo configuration (left); RoboNova servo-bracket configuration (right).

SERVO 11.2007

15

MrRoboto.qxd

10/9/2007

10:35 AM

Page 16

In Figure 3, you may have noticed that wires to the servo are all black, and not the traditional yellow-red-black colored wires in other Hitec servos. This can make determining the signal wire from the ground wire a bit challenging. The signal wire is actually dark-gray in color, but it can be difficult to see. The signal wire is the wire closest to the two tabs on the side of the connector housing (see Figure 4). The power wire is in Figure 4. Closeup view of the connector for the center, and the ground the HSR-8498HB servo — the control signal wire wire is on the other side of is dark gray in color located next to the side tabs the connector. on the connector. One of the advanced features of these servos is configured to look like a standard that they can also be controlled and R/C servo, and the other servo is programmed via RS-232 serial configured for a robotic knee joint. communications. In fact, 127 of these Each servo comes with an accessory servos can be daisy-chained together package with the different configuraon one signal line. In addition, the tion options. Table 1 shows the serial communication protocol can specifications for this servo. allow changing the proportional and

derivative gains and dead bands on the servo, servo ID, battery status, position, current draw, and turning the servo on and off. The servos that I used for this answer had an older firmware version which doesn’t allow for the serial communication. The minimum firmware revision must be at least v1.10. There is a little sticker on the inside of the servo’s bottom plate which identifies the firmware version. I have sent my servos back to Hitec to have the firmware version upgraded. Next month, I will continue this topic with a discussion on how to use serial communications to control this servo. For more details about the PWM position feedback signals and the serial communication protocol, go to the Hitec Robotics website (www. hitecrobotics.com) and look under the download page for the “Pulse of HMI Protocol” subject. There you will find a file called “HMIprotocol.pdf,” that will keep you busy until next month. SV

HE8EJIFH;<;H>?J;9)0' .EW2OBOT3ERVOS >_j[Y Ej^[h8hWdZ

+656*

+656*

+657*

6SHHGVHF 7RUTXHR]LQ 6WHHO*HDUV

6SHHGVHF 7RUTXHR]LQ 6WHHO*HDUV

6SHHGVHF 7RUTXHR]LQ 7LWDQLXP

$//63(&,),&$7,216$792/76

7KHUHVXOWVRIDQLQIRUPDOSROOWDNHQUHFHQWO\DWWKH)LUVW $QQXDO:RUOG'RPLQDWLRQ6\PSRVLXPDUHQRZLQ5RERWV SUHIHU+LWHFVHUYRVRYHURWKHUVHUYREUDQGV7KH\NQRZWKHZLGH VHOHFWLRQRI+LWHFDQDORJDQGGLJLWDOVHUYRVSURYLGHWKHPZLWKWKHSRZHUDQGGHSHQG DELOLW\QHHGHGWRHYHQWXDOO\WDNHRYHUWKH:RUOG0DNH\RXUURERWKDSS\XVH+LWHFVHUYRV 3DLQH6WUHHW_3RZD\_&DOLIRUQLD___ZZZKLWHFUFGFRP

16

SERVO 11.2007

Full Page.qxd

10/10/2007

11:42 AM

Page 17

Events.qxd

10/9/2007

11:28 AM

Page 18

Send updates, new listings, corrections, complaints, and suggestions to: [email protected] or FAX 972-404-0269 Know of any robot competitions I’ve missed? Is your local school or robot group planning a contest? Send an email to [email protected] and tell me about it. Be sure to include the date and location of your contest. If you have a website with contest info, send along the URL as well, so we can tell everyone else about it. For last-minute updates and changes, you can always find the most recent version of the Robot Competition FAQ at Robots.net: http://robots.net/rcfaq.html

Oahu, HI ROVs built by university and high school students compete in this event, which is part of the MATE (Marine Advanced Technology Education) series of contests. www.marinetech.org/rov_competition

24

ROBOEXOTICA Museumsquartier, Vienna, Austria A competition for “cocktail robots” that includes events such as serving cocktails, mixing cocktails, bartending conversation, and lighting cigarettes. www.roboexotica.org/en/acra.htm

24

Roaming Robots Grand Final Kent, UK Remote-controlled vehicles destroy each other. www.roamingrobots.co.uk/events_calendar.htm

30

ARADBOT Santiago, Dominican Republic Autonomous robots compete in line-following and mini-Sumo events. www.aiolosrd.com

30

Texas BEST Competition Texas Tech University, Lubbock, TX Students and corporate sponsors build robots from standardized kits and compete in a challenge that changes each year. www.texasbest.org

— R. Steven Rainwater

N ov e m b e r 3

DARPA Urban Challenge The former George Air Force Base, Victorville, CA Autonomous ground robots compete against each other in a simulated urban environment to complete a waypoint-following course. www.darpa.mil/grandchallenge

10-11 Canadian National Robot Games Ontario Science Center, Toronto, Ontario, Canada Events include novice, advanced, and master mini-Sumo, full-size autonomous and RC Sumo, fire-fighting, line-following, photovore, a walking robot race, and a search and rescue contest. http://robotgames.ca

16-17 All Japan MicroMouse Contest Tsukuba International Conference Center, Tsukuba, Japan Includes Micromouse, Micromouse Expert level, and Micro Clipper events. www.robomedia.org/directory/jp/game/ mm_japan.html

17

DPRG RoboRama Museum of Nature and Science, Dallas, TX The usual assortment of events including Quick Trip, T-Time, wall-following, line-following, and can retrieval. An outdoor waypoint-following event known as the Long-Haul will also be included this year. www.dprg.org

23-24 Hawaii Underwater Robot Challenge UH Manoa Duke Kahanamoku Aquatic Complex,

18

SERVO 11.2007

December 7-8

South’s BEST Competition Beard-Eaves Memorial Coliseum, Auburn University, Auburn, AL Regional BEST teams from multiple states compete in this regional championship. www.southsbest.org

8

Penn State Abington Robo-Hoop Penn State Abington, Abington, PA Autonomous robot basketball event in which robots must pick up foam balls and shoot or dunk them into a basket. www.ecsel.psu.edu/~avanzato/robots/con tests/robo-hoops

Full Page.qxd

10/10/2007

11:44 AM

Page 19

Nov07NewProd.qxd

10/10/2007

4:56 PM

Page 20

New Products

N E W P RO D U C T S CONTROLLERS & PROCESSORS 2 MHz Pulse-Width Modulator

M

icrochip Technology, Inc., now offers the MCP1631 2.0 MHz, highspeed pulse width modulator (PWM). The highly integrated device contains a 1 ampere integrated MOSFET driver, high-speed comparator for over-voltage protection, and both battery-current and voltage-sense amplifiers in one small package. Protection features, such as under voltage lock out (UVLO) and over-temperature protection, come standard with the PWM, which is capable of charging multiple battery chemistries, including Li-Ion, NiHM, NiCd, and Lead Acid. The MCP1631 PWM provides a means to close the

feedback loop in Switch-Mode Power Supplies (SMPSs) that use microcontrollers for general system intelligence and control. The integration of the PWM’s SMPS input and output interface via its voltage comparator, batteryvoltage, and current-sense amplifiers — plus its 1A MOSFET driver — enable designers to use this single device to perform many different functions in their designs. The result is a smaller design footprint and lower overall cost. Additionally, the MCP1631 is controlled by an easilyprogrammable microcontroller, meaning that exact charge profiles for a variety of battery-charging systems can be accurately met, while closing the feedback loop with the speed and precision needed for safe charging and long battery life. The PWM’s UVLO and over-temperature protection features enhance the safety of battery-charger designs. High-voltage versions of the MCP1631 PWM are available (Part # MCP1631HV), which operate from 6V to 16V and include a linear regulator (LDO). The standard versions of the device operate from 3V to 5.5V, and do not include a LDO. Possible applications include handheld

Robot Controllers

Programmable Robot Kits

Wiring Robot Controller

INEX MicroCamp Mega8

· Atmel ATMega 128 · 128k Memory · 43 Digital I/O Pins · 8 Analog Inputs · 8 External Interupts · 6 PWM Channels · 2 Serial Ports including Bi-Directional USB · The Wiring Programming Language

· Atmel ATMega8 · Dual DC motor drivers · 2 Buttons, 2 LEDs · Serial port · 5-Analog ports for sensors · +5V switching power supply · No soldering required · Supports In-system Programming via ISP connector with included PX-400 Serial Programmer

The Wiring language provides a simplified subset of C or C++ that hides more advanced concepts like classes, objects, pointers (while still making them accessible for advanced users). You get the power of C or C++ with the ease of a language like Basic. Programs execute at full C++ speed on the board.

Includes eveything you need to build a simple mobile robot. Add your own additional sensors for even more complex robots.

$59.95

$69.95

MicroBric Viper

ARC1.1 Robot Controller • Atmel ATMega16 • 1k SRAM, 16k Flash • Dual 1.1 amp motor drives • Supports motors up to 25V • Dual quadrature encoder support • Programming cable included with kit • No additional hardware needed • Works with BASCOM and AvrDude programming software Ideal for controlling your small robot. With a Microcontroller and onboard motor controllers, you get all the electronics that you need (except sensors) on one board. Kit $37.95

/ Assembled $41.95

Also Available: Electronic Components Servos Motors Hardware Wheels & Tires and More! More New Products on the way!

· Screw-together Assembly · BasicAtom Microcontroller · 2 motor modules · Bump sensor modules · Switch Modules · IR Remote & Receiver Module With microbric, you can build complex electronic devices with little or no prior electronics knowledge. As no soldering is involved and the parts are fully reusable, you can build and rebuild programmable robots as many times as you like.

$89.95

1-800-979-9130

MaximumRobotics.com 20

SERVO 11.2007

Nov07NewProd.qxd

10/10/2007

4:57 PM

Page 21

New Products medical, consumer, and industrial electronic devices that require power management and SMPS technology, with a focus on battery charging. Examples include intelligent power supplies, smart battery chargers, RF remote devices, handheld scanners, parallel power supplies, and AC power factor correction. For further information, please contact:

Microchip

Website: www.microchip.com

MOTOR CONTROLLERS Robot Controller Board That Does it All

E

fficient Computer Systems, LLC announces the BOTLOGIC Controller. This 4 x 6 inch board controls up to 32 RC servos; 24 of these servo channels have load sensing circuitry to allow robots to detect the amount of force being applied by each servo. This servo feedback will help a robot sense when a leg has touched the ground, as well as how much of the robot’s weight is

on each leg. When used with a gripper, it will help detect when the gripper has touched an object and how much force the gripper is actually applying to an object. Twenty user inputs are available for connecting to bumper switches or other sensors, allowing your robot to explore its environment. Also on-board is a threeaxis accelerometer which is perfect for today’s balancing BOTs. The built-in SD card interface can be used for loading new programs into the robot, as well as storing data and sound files to use with the built-in audio recording and playback circuitry. User messages or diagnostic data can be displayed on the 2 x 16 character LCD. High current LED drivers can deliver power to up to six externally mounted high brightness LEDs to illuminate the environment, show system status, or just look cool. The three two-amp solenoid drivers can be used to power accessories such as motors or fans. The daughterboard connector allows for future expansion or custom add-on features. Control your robot with the optional wireless interface to a PC or through most wireless/wired Playstation 2 Gamepad controllers. Development tools are available for both Basic and C. For further information, please contact:

Efficient Computer Systems, LLC

Website: www.BOTdeveloper.com

SERVO 11.2007

21

CombatZone.qxd

10/9/2007

10:28 AM

Page 22

Featured This Month Participation 22 Being a Safety Jerk by Kevin Berry

23 Rules for First Time Participants

by Kevin Berry

Feature 23 Updating the British Fleet by James Baker

Technical Knowledge 26 Building a Lightweight Launch Bot

by Chad New

Events 25 Results — Aug 11 - Sep 11 27 Upcoming — Nov & Dec

22

SERVO 11.2007

The December edition of Combat Zone will be “All Armor.” We solicited articles from builders across the spectrum, from antweights to superheavies, from newer builders to seasoned pros. Like our “Welding Special” and “Heavy Power” editions, this is intended to be a resource for builders who want to create tougher, nastier bots. Which is really what the sport is all about! — Kevin Berry

PARTICIPATI N Being a Safety Jerk ● by Kevin Berry

I

n my opinion, there is no tougher job at an event than being the Safety Officer (except for Event Organizer, that is). Most of the builders are your friends, they’ve worked hard to get ready, and are very possessive of their designs. You — as the person in charge of keeping people from getting hurt — have a real responsibility to do the right thing. The first step when asked to be a Safety Officer is to have a discussion with the EO. You need to understand their tolerance for “gray areas” in

the rules. For example, if the rule says a bot must “spin down” in 30 seconds, what do you do with a “33 second” bot? Do you insist they meet the 30 seconds, are you allowed to use judgement, or do you refer this to the EO? Understanding the rules of engagement is an important step in avoiding misunderstandings. The next step is to communicate with the participants ahead of time, if possible. Let them know you are going to be tough on enforcement, so they aren’t surprised during check-in. EOs vary widely on safety, so someone coming from an “easy” event to a

CombatZone.qxd

10/9/2007

10:28 AM

Page 23

“tough” one might be shocked or angry if they are denied entry. The final step during actual safety checks is to make sure to be calm, level-headed, and non-involved. That means making the rules the bad guy, not the Safety Officer. “I’d like to overlook that weapon twitch, but the rules say ‘no movement,’ and it’s underlined, so you’re going to have

to fix it.” Another tack is personal, but not threatening. “If you or someone else gets hurt, I’m going to feel terrible about it, even if you say it’s not my fault. Let’s work together to fix that.” Sooner or later, though, you are going to have to “86” a bot. Before I do that, I consult with the EO, even if we have reached an earlier

understanding. Good will is 90% of running an event, and spending the EO’s good will ought to involve them. In the long run, out of the frenzy and emotion of an event, it’s easy to see that if the rules are known, builders should follow them, and our sport prides itself on a clean safety record. No builders, staff, or spectators hurt. Period. SV

Rules for First Time Participants ● by Kevin Berry

S

tarting in any new sport, club, or activity brings with it some uncertainty, hesitancy, or just plain lack of knowledge. Robot combat is certainly the same way, especially when coupled with the concept of instant destruction of your hard-built, first creation! Like anything in life, the best approach is a combination of humble demeanor and regular, thoughtful questioning. Veteran builders represent a cross section of society: from casual participants to intense competitors; tolerant to impatient; non-technical to deeply specialized. No matter how tolerant or patient, nobody wants to reply to a post or email like this one: “I want to build a robot and I don’t know how. What

do I do?” The first answer to this type of question is “do the research!” There are several good books, many team and club websites, and locators for nearby competitions that are available to first time builders. Next, lay out a design. Even if you aren’t sure exactly what you are doing, give it a try. Builders are much more likely to give advice if it’s in context of a potentially buildable bot, rather than a mythical dream machine. “I laid out this bot using the Whyachi gearboxes, but I really can’t afford them. Anyone got experience with less expensive alternatives like modified HF drill motors?” This kind of question will be much better received and — more importantly — get a better answer.

Builders are busy people and really don’t want to spend a lot of time educating people who won’t do their homework. Finally, try to meet some builders in person. Sure, we’d all like to take a bot to our first event, but maybe a better approach is to go as a volunteer, get access to the pits, spend time with participants, and learn, learn, learn! There has never been an event held that has had enough willing hands, skilled or not. I know after getting eliminated in the very first fight of a three day event, my son and I spent the next two sweeping the box and wrangling bots, and had a hugely successful introduction to the sport. SV

UPDATING THE BRITISH FLEET Evolution of a Rob t Army ● by James Baker

A

s in the USA, the sport of robot combat lives on in the UK, despite the continued absence of newly televised events on the scale of Robot Wars and Battlebots. This sport survives only for the efforts of those few dedicated event organizers. On both sides of the Atlantic,

these events sustain robot combat. In turn, it is the robot builder and their machines that sustain the event organizers and their A selection of robots from the Xbotz fleet. The current update aims to use just two speed controller types across all of the robots.

SERVO 11.2007

23

CombatZone.qxd

10/9/2007

11:15 AM

Page 24

“Wheely Big Cheese” climbs the fence at a school science fair.

The 30 lb robot “Bug” with electric grabber, uses a Scorpion XL for drive and an Mtroniks Viper for the grabber.

events. The more combat robots available to event organizers, the stronger our sport becomes. My teammates and I (Team Xbotz) keep a robot battle fleet bigger than most, but as in the military, even the biggest fleet needs to be kept up-to-date and at the sharp edge to survive. The Xbotz fleet has been fighting regularly for a couple of years now, and to be honest, are starting to look very tired. The technology they use is now seven years old in some cases, so a major Our 220 lb robot “Wheely Big Cheese” with 250 lb driver James Baker. This robot lost five lbs just by swapping to a Sidewinder.

update of all the robots is now being undertaken. This article hopes to give a little insight into the efforts of one team to stay competitive, but also practical, as running an active fleet this big is a huge drain on time and resources, but I personally think it is really worth it. We are still building new machines, as well. For whatever reason, be it arena configuration, available components, or a wider statement of national stereotypes, the past has shown that the majority of British robots lacked the horsepower of their American cousins, relying instead on good control and agility to bring their weapons to bear, but leaving them visibly slower. By far, the most popular drive solution for robot combat in the UK was the Bosch 750 watt, 24V motor with 4QD speed controllers. This was a good solution for us, and it was rare to see someone successfully use more powerful drive systems. All of the currently active Xbotz heavyweight robots originally used 4QD speed controllers. It is a big job to completely overhaul a heavyweight fighting robot, but to update a whole fleet The 30 lb robot “Tantrum” uses NiMH batteries and a Scorpion XL for drive, with separate NiCd cells for the weapon motor.

24

SERVO 11.2007

Our 60 lb walking artbot “Venom” with electric crushing pincers uses a single Robot Power Scorpion XL and R/C switchers.

of heavyweights, featherweights, hobby weights, and artbots within a realistic timescale and budget required a very clear decision-making process. We decided that the upgrades must all be reliable; easy to use; offer a packaging, weight, or size advantage; and be good value for the money. We also decided that logistically, having all the robots on common components where possible would be a huge advantage. This is mainly so we can carry fewer spares. First upgrades on the list were chassis, armor, and weapons, but those subjects are a whole other article, so we can skip those for now. Second on the list were drive systems — namely motors, gearboxes, batteries, and speed controllers. The wheels are fine as they are. All the robots have good, wellproven drive systems already, so rather than change the motors and gearboxes, it made more sense to us to update the speed controllers and batteries only. Speed controllers are one area of your robot you must get right, or you will always struggle in the arena. We did not want to take advice and Our other 220 lb robot “Edgehog” has swapped 4QD speed control and Hawker batteries for a Robot Power Sidewinder and NiCd cells, saving over 20 lbs. This was used for extra armor.

CombatZone.qxd

10/9/2007

10:28 AM

Page 25

Another 220 lb robot “Carnage” uses a Robot Power Sidewinder for both drive motors and another for both weapon motors. The Lead Acid batteries have been replaced with NiCd packs, saving 10 lbs.

then regret it later, so we tried several controllers’ head to head. I was so impressed with one particular speed controller; I became the European agent for the company that makes them. The Robot Power Sidewinder looked really good on paper, and during testing it exceeded our requirements by a long way, so we now use them in all of our heavyweights and one of the featherweights. We had a similar experience with the smaller bots, identifying the Robot Power Scorpion XL as — by far — the best controller for all of the sub-30 lb machines, and the artbots, except one. We used a Robot Power Scorpion XXL (a modified XL) for this one robot, as it needed the extra power the XXL offers. The Robot Power Sidewinder was the mainstay of the UK championships this year, stepping in to keep many robots running when their regular controllers failed. One key feature to this was that the speed controller being removed from a robot was always bigger than the Robot Power option, so it always fit. The space saved in our

A Scorpion XL is at the heart (literally) of our 30 lb artbot “Hellraiser.”

axe-bot Edgehog was unbelievable; swapping a laptop-sized electronics box for something so small gave us room to make much needed weapon modifications. Batteries are also an area where you cannot afford to get it wrong or your robot is going to struggle. All Xbotz heavyweights ran Hawker sealed lead acid batteries prior to the overhaul, which always served us well, but a switch to nickel cadmium cells was an easy upgrade, allowing us to lose weight and gain voltage. Two of the smaller bots now use these also, with the remainder of the fleet running nickel metal hydride cells. We considered lithium based power, but decided on the NiMH mainly because of the lower cost. Our third area to look at was radio control. All Xbotz robots use 40 MHz Futaba systems, and we

saw no reason to change that. The new 2.4 GHz technology is very impressive, but the 40 MHz equipment has always done the job for us, so the money was better spent elsewhere. The updates to the fleet are progressing well, with the robots done to date showing improved speed and agility. Wheely Big Cheese, for example, is completely transformed. The whole Xbotz army of robots will be updated before the 2008 season starts, with detailed information about each robot, and each event we attend available on our website (www.xbotz.com) plus an online shop for European builders, thanks to www.leafish.co.uk. See www.robotcombat.com or www.robotpower.com for information about the speed controllers we use at Team Xbotz. SV

EVENTS RESULTS — August 11th - September 11th

R

obot Battles held their annual Labor Day event at Dragon*con in Atlanta, GA. Results were not available at press time, but will be featured in an upcoming issue. Go to www.robotbattles.com for more information.

28th-31st at Haven Hafan y Môr Holiday Park, North Wales. Go to www.roamingrobots.co.uk for more information. Results are as follows:

T

● Heavyweight — 1st: Big Nipper; 2nd: Terrorhurtz; 3rd: Iron Awe 5. ● Annihilator — Tilly Ewe 2.

he UK Heavyweight Championships, presented by Roaming Robots, were held August

● Featherweight — Little Flipper 2.

R

obots Live presented a show on September 1st-2nd at Sportspace Hemel Hempstead. Team Wind Power took home 1st place in the Feathers and Heavies, and Team Big Nipper took home 2nd in both weight classes. Go to www.robotslive.co.uk for more information. SV

SERVO 11.2007

25

CombatZone.qxd

10/9/2007

10:29 AM

Page 26

TECHNICAL KN WLEDGE Building a Lightweight Launch Bot ● by Chad New

T

hese days, it seems that the trend within combat robotics is for people to build two basic types of robots. They are either a super powerful KE spinner which tries to destroy everything in its path or super tough wedge type robots made to withstand the punishment from the spinning robots. Some robots will deviate from this trend, but for the most part, the majority fall into these categories. In the half dozen years that I have been involved with this sport, I have mainly built sub-light robots, which are robots that are less than 30 pounds. I have always wanted to build a pneumatic robot capable of launching another robot into the roof; however, due to the weight restrictions on the sub-light classes, I was never able to build a robot that could achieve my goal. One day after seeing videos

of various UK pneumatic robots, I decided it was time to build a full size 60 pound lightweight robot. This robot would be something different than the norm; a robot that could (hopefully) take the abuse from the powerful KE robots and send the wedge type robots flying out of the arena. In this build report, I will give a description of how my 60 pound ‘launch bot’ “Rocket” became a reality.

The Design

In today’s game, there are many robots that wield weapons that can destroy an opponent in short order, especially in the lightweight class. For this reason, I enlisted the help of my friend Bryan Ruddy to help me CAD and design this robot. After much talking, we agreed on a list of things that had to be achieved for this robot to PHOTO 1. After many weeks of design, be successful. the final CAD emerged. We both agreed that the robot should be centered on the flipping system, be highly maneuverable, have an armor system which could stand up to the tough KE robots, and — most importantly — be able to shoot an opponent to the roof! After a few weeks of CADing, this is what we came up with (Photo 1).

Drive The drive on Rocket uses four BaneBot 42 mm 16:1 (www.banebots. com) gear motors upgraded with the 775 PHOTO 2. Parts cut and systems installed, the frame is now ready for wiring and final assembly.

26

SERVO 11.2007

sized motor. This drive package mounted to 3” wheels gives a great balance of speed and pushing power. Also, the placement of the motors and shape of the base plate were designed for optimal maneuvering and to eliminate scrubbing of the wheels. When mounted correctly, I have found the BaneBot units to be very durable and work extremely well.

Armor Other than the pneumatic system, the armor arrangement is my favorite part of this robot. It was designed to take and repel the attacks from other robots, as well as facilitate self-righting by letting the arm contact the ground at all times. All of the sides are sloped in order to reduce the amount of surface area that other robots might be able to get a hit on. The armor is very low to the ground making it difficult for vertical spinning robots to grip and damage. Also, the armor is mounted to the base plate on a system of custom rubberized shock mounts which allow the armor shell to move and somewhat flex when hit. Made from 1/8” titanium, the shell is extremely durable and has also been designed to allow various attachments to be added, depending on the opposing robot; all of which have yet to be debuted.

Electronics Rocket utilizes some of the best electronics available in order to free up weight and improve durability. Control is handled by the new Airtronics M11 2.4 GHz system (www.airtronics.net) which they

CombatZone.qxd

10/9/2007

2:44 PM

Page 27

modified for me by adding a button which can be pressed when I want the flipping arm to be activated. This button allows me to focus on driving the robot while giving the flipping control to a team mate. The battery is a 5,000 mAh 14.8V Li-Poly made by Thunder Power (www.thun PHOTO 3. Rocket putting the derpowerrc.com). This pack LW wedge ‘Homer’ into orbit. was able to free up a lot of weight, allowing a better distribution paintball tanks, also with an inline for other items. The speed control is buffer tank to allow for gas expanRobot Power Sidewinder; which may sion. This arrangement gives Rocket be overkill; however, it gets the job over 20 shots at getting its opponent done well. The guys down at out of the arena. With the massive BaneBots built and designed me a amount of force that this system custom voltage booster/timed switch produces, the arms were made from to operate the valve for the cylinder. .6” titanium with a .5” S7 steel pushThe valve requires very high voltage ing plate. Overall, I believe this to be one of the best and toughest flipping and only needs to be open for a systems in the lightweight class. fraction of a second; this device takes care of that task.

PHOTO 4. The armor mocked up onto the frame ready to be sent for welding.

robot able to right itself. I believe that once I fix this problem, Rocket will be a serious contender in the lightweight class. I also hope that by building an exciting pneumatic robot, it will spur interest in other types of robots, not just the powerful KE spinners and brick wedges. I strongly encourage everyone to go after their goals because you never know what might come out of them! SV

Conclusion

Flipper The flipping system uses a custom-made pneumatic cylinder capable of withstanding C02 at pressures greater than 800 PSI. Activation of this cylinder is controlled by a large solenoid valve. Feeding this system is two 20 oz

After Rocket’s first event at RoboGames 2007, I found a flaw in the design where the armor will rub against the arm when inverted, thus not allowing the arm to retract into the body making the PHOTO 5. Shined up and with new stickers, Rocket, ready for its first event.

EVENTS UPCOMING — November and December

H

ORD Fall 2007 will be presented by the Ohio Robot Club in Brecksville, OH on November 3rd. This event is for Fairy, Ant, and Beetle weight combat robots. It will be held at the Cuyahoga Valley Career Center (CVCC) (south east of Cleveland). For complete details including rules, safety forms, release forms, maps, and local hotels, see

their website at www.ohiorobot club.org.

R

oaming Robots will present an event at the Maidstone Leisure Centre in Kent, England, on November 24th, and at the Harvey

Hadden Sports Complex December 1st in Nottingham, England. Go to www.roamingrobots.co.uk for more details.

R

oboChallenge will present their Thinktank Christmas Special December 28th-29th in Birmingham, England. Go to www.robochalle nge.co.uk for more details. SV

SERVO 11.2007

27

Cieslak.qxd

10/9/2007

11:25 AM

Page 28

AUTOFLEX 2.0 New and Improved Autonomous Programming Tool for FIRST Robots by Brian Cieslak

I

n the March ‘06 issue of SERVO Magazine, I introduced you to a program called AutoFlex — a tool used for developing autonomous routines for FRC (First Robotics Competition) robot controllers.

The program was created by members of FRC Team 1675 when they realized that they were going to the FIRST (For Inspiration and Recognition of Science and Technology) National Championships in Atlanta, GA without any autonomous functionality for their robot. Without having access to the robot until the event, they needed a way to quickly program the robot to perform some task during the autonomous period. The solution was to create a program that would allow the team’s driver to teach the robot what it had to do during the autonomous period by recording the driver’s commands as he drove through the autonomous routine. Training took place on the practice field before the matches started. At the beginning of the match, the robot would repeat the commands that it was taught. In 2005 during the Triple play competition, the robot scored two tetras during each autonomous period. During the 2006 Aim High competition, the robot could drive Photos courtesy of FRC Team 1675 — The Ultimate Protection Squad.

28

SERVO 11.2007

up to the goal and shoot an entire magazine of balls through the hole (most of the time). The original program was a little cumbersome and complicated to use. While the driver commanded the robot through the routine that was to be recorded and then played it back during the autonomous mode, a programmer — with a laptop connected to the robot via a serial cable — chased (or was chased by) the robot as he captured data. The data then had to be loaded into a file and the whole program was recompiled and reloaded into the robot. From the sidelines, this was fun to watch, but those actually involved in the process were often quite stressed and in peril. Autoflex has been simplified and updated to version 2.0 to take advantage of the internal EEPROM memory available in the FRC robot controller. Commands are now written directly to the EEPROM memory. No

FIGURE 1. Team 1675’s first robot programmed with Autoflex for the FIRST Triple Play competition would know one tetra from the goal and cap the second during the autonomous period.

more laptops and cables, editing data, and reprogramming. Programmer stress levels have been greatly reduced!

Imagine This! During practice, you set your robot on the playing field, click a button on the operator interface and start driving. Then you set your robot back to the starting point, connect a dongle to the competition port of the operator interface and flip the dongle switch to autonomous and the robot will replay the practice session you just recorded. Don’t like what you see? Just reset the dongle switch back to the off position and just click the program button again to re-record another session until you get it right. Neat, eh?? Your robot is

Cieslak.qxd

10/9/2007

11:25 AM

Page 29

ready to run during the autonomous period, all in about 15 minutes.

Now you are an expert.

Getting Started

For the Beginning FIRST Programmer

A .zip file can be downloaded from the FRC Team 1675 website (see Web Links sidebar) that includes a version of the FRC default code with AutoFlex included. I/O mapping for the default program is as follows: Joystick Joystick Joystick Joystick

1 2 1 2

-

Y axis to PWM_1 Y axis to PWM_2 X axis to PWM_5 X axis to PWM_6

Set your robot to program mode and download the FRCAutoFlex.hex file using the IFIdownloader program available from the IFI website (www. ifirobotics.com). Attach a programming dongle to the competition port of the operator interface (instructions on how to make your own are also available from the IFI website) and set the autonomous switch to the open position. You are now ready to start programming your robot for autonomous operation. The FRCAutoFlexCode.hex program records four inputs: joystick 1 x-axis, joystick 1 y-axis, joystick 2 x-axis, and joystick 2 y-axis. Click the trigger on the port 1 joystick to start recording. You now have 15 seconds to drive through your autonomous routine. After 15 seconds, the robot stops recording commands even though it lets you keep driving. To replay what you just recorded, ‘close’ the autonomous switch on the dongle. Watch out! Your robot will start to execute the code you just recorded. The robot will play 15 seconds of commands and then stop until you open the autonomous switch again. Once you are satisfied with the autonomous routine you’ve recorded, place a jumper on the ‘digital input 1’ pins. This write protects your autonomous program from being accidentally erased if you click the trigger while driving around. That’s the basic operation.

If you are just learning to program a FIRST robot, a sample project that is fully functional is included in the zip file you can download from the Team 1675 website that can serve as a template to get you started. The programming kit that comes with your robot includes a disk with the MPLAB-IDE programming environment and the C18-Complier Version 2.4, as well as the downloader program. You will need these tools to compile and download your program to the robot.

Adding Autoflex to Your Existing Code Adding AutoFlex to your existing code is simple if all the calls to your control functions (motor control, manipulator arm, etc.) are made from the Default_Routine() function found in the User_Routines.c file. You must do the following (refer to the sample code provided):

FIGURE 2

shown in Figure 2. Also add the #include”AutoFlex.h statement at the beginning of the file. 3) Open the user_routines.c file. Add a call to the function autoflex_recorder() to the Process_Data_From_ Master_ uP() function as shown in Figure 3. Also add the #include”AutoFlex.h” statement at the beginning of the file. 4) Open the main.c file. Add a call to the function rewind_autoflex_ playback() to the main()function as shown in Figure 5. Also add the #include"AutoFlex.h" statement at the beginning of the file. FIGURE 4. Team 1675’s Aim High robot would drive up to the goal and shoot most of its 10 balls through the hole.

1) Copy the following files to your project folder, then open MPLAB and add them to your project: a) AutoFlex.c b) Autoflex.h c VEX_eeprom.c d) VEX_eeprom.h 2) Open the user_routines_fast.c file. Add a call to the function autoflex_playback() to the user_ autonomous_code() function as

FIGURE 3

SERVO 11.2007

29

Cieslak.qxd

10/9/2007

11:25 AM

Page 30

Web Links For Autoflex files http://team1675.com/ teamdownload.html

FIGURE 5

5) Configure Autoflex.h to reflect your robot system. Sections that you may want to consider changing include the following: a) Determine how many inputs you want to capture and which ones. //add defines here to assign //commands to user controls that // you want record/ #define AUTO_COMMAND1 p1_x //left Joystick x #define AUTO_COMMAND2 p1_y //left Joystick y #define AUTO_COMMAND3 p2_y //right joystick y #define AUTO_COMMAND4 p2_x //right joystick x //#define AUTO_COMMAND5 //uncomment to add another input //#define AUTO_COMMAND6 //uncomment to add another input // Number of inputs we plan to //record // Default is set up to save 4 inputs. //You can save up to 6 // inputs. You can define two auto //command lines above. // then change the number of FIGURE 6. Autoflex was used to program a large claw-like manipulator during the autonomous period at the beginning of the Rack-n-Roll competiton.

For competition port dongle www.ifirobotics.com/ oi.shtml

//inputs on the line below. #define NUM_OF_INPUTS 4i b) You can determine what you want to use as the ‘Record Button.’ The default is port 1 trigger Button. //define the mechanism that will //act as the record button. //In this example port trigger is a //button on the OI. // that you would press to the //forward position to start recording #define \ AUTO_BUTTON_REV_THRESH \ (unsigned char)100 // used by Vex #define \ AUTO_BUTTON_FWD_THRESH \ (unsigned char)154 // used by Vex #define \ AUTO_NEUTRAL_PWM_VALUE \ (unsigned char)127 #define AUTO_RECORD (p1_sw_trig) //port_1 trigger to start recording c) You can adjust the length of time you want to record commands by changing the TIME_LIMIT value. Default is 150 tenths of a second (or 15 seconds). The maximum value of TIME_LIMIT depends on the number of inputs you are trying to save. The max number of command values that can be saved is 1,024. To determine the max time available, use the following formula

Contact the Author Brian Cieslak is a mentor for FIRST Team 1675, The Ultimate Protection Squad. He can be contacted via email at [email protected].

30

SERVO 11.2007

(1024/ number_of_inputs) - 1 = max_tenths_of_seconds. For example: (1024/4 inputs)-1 = 255, so TIME_LIMIT could be set to 255 tenths_of_seconds. (25.5 seconds). // The length of the autonomous routine in tenths of seconds #define TIME_LIMIT 150 d) You can assign which digital port you want to use for your WRITE_ PROTECT jumper. If you don’t want to write protect your autonomous code or you have used up all your digital ports, re-define WRITE_ PROTECT to ‘1.’ // if jumper in place then do not //record (assuming jumper pulls pin //low) #define WRITE_PROTECT \ (rc_dig_in01) e) Since an FRC robot uses a longer timing interval than VEX robots during autonomous operation, uncomment the #define FRC 1 line to adjust the timing if you are adding Autoflex to a FRC robot.

No More Excuses to Sit Idle! When I attended FIRST Regional competitions in Milwaukee, WI and Cleveland, OH and the FIRST National Championship in Atlanta, I was surprised by how many robots sat idle during the autonomous segment of the match. Our team started touting the benefits and simplicity of the Autoflex program there and enabled several teams to compete during that 15 second period at the beginning of the match. Even sending the robot out to a defensive position is better than just sitting there. I do want to emphasize, though, that Autoflex is not a substitute for a well thought out autonomous program that uses sensors and feedback algorithms. To be truly autonomous, the robot must be aware of and react to its environment. So programmers, you are not off the hook. See you in the ‘Pits.’ SV

10/10/2007

6:15 PM

Page 31

I M S e r v ic e

P .O . B o x 1 4 2 , N e w H u d s o n , M I 4 8 1 6 5

V o ic e : 8 8 8 -4 5 1 -1 6 7 0 o r : 2 4 8 -4 8 6 -3 6 0 0

D e s k to p C N C S o lu tio n s 1 2 x 1 2 x 4 9 0 o r 1 7 5 ip m 1 2 x 2 4 x 4 N E W 4 0 0 ip m S e r v o C N C R o u te r s S ta r tin g a t $ 1 9 9 5 D e s k C N C

7KH(VFDSH 5RERW¶VEXLOWLQ PLFURSURFHVVRU HQDEOHVLWWR ³WKLQN´RQLWVRZQ .65 

6 UH ROGH TX UL LUH QJ G 

ShowcaseNov07.qxd

7KHURERWIURJPRYHVIRUZDUG ZKHQLWGHWHFWVVRXQGDQG UHSHDWVVWDUWPRYHIRUZDUG !

VWRS!OHIWWXUQ!VWRS!ULJKW WXUQ!VWRS.65 

PP:KLWH ZDWHUFOHDU /('9 PFG $% 

 VHFRQGYRLFHUHFRUGHUSOD\EDFN PRGXOH7KHHOHFWUHWPLFURSKRQHLVRQWKH ERDUG2QHEXWWRQUHFRUGVWKHRWKHUEXWWRQLV PRPHQWDULO\SUHVVHGWRUHSOD\WKHPHVVDJH SUHDVVHPEOHG $ 

i m s r v .c o m

7KH9HOOHPDQ3HUVRQDO6FRSH LVQRWDJUDSKLFDOPXOWLPHWHU EXWDFRPSOHWHSRUWDEOH RVFLOORVFRSHDWWKHVL]HDQG FRVWRIDJRRGPXOWLPHWHU +36  QGV 7KRXVD VRQ P WH L PRUH LWH V E RXUZH

Robotics Showcase

THOUSANDS OF ELECTRONIC

PA R T S A N D S U P P L I E S

VISIT OUR ONLINE STORE AT

The Power SchmartModule™ will power up your circuits with your choice of -9, -12, +2.5, +3.3, +5, +9, and +12 volts.

www.allelectronics.com WALL TRANSFORMERS, ALARMS, FUSES, CABLE TIES, RELAYS, OPTO ELECTRONICS, KNOBS, VIDEO ACCESSORIES, SIRENS, SOLDER ACCESSORIES, MOTORS, DIODES, HEAT SINKS, CAPACITORS, CHOKES, TOOLS, FASTENERS, TERMINAL STRIPS, CRIMP CONNECTORS, L.E.D.S., DISPLAYS, FANS, BREADBOARDS, RESISTORS, SOLAR CELLS, BUZZERS, BATTERIES, MAGNETS, CAMERAS, DC-DC CONVERTERS, HEADPHONES, LAMPS, PANEL METERS, SWITCHES, SPEAKERS, PELTIER DEVICES, and much more....

$15

O R D E R TO L L F R E E

Very Schmart!

1-800-826-5432

www.schmartboard.com

Ask for our FREE 96 page catalog

SERVO 11.2007

31

Duesing.qxd

10/10/2007

12:33 PM

Page 32

A look inside the mechanics of combat robots by Brett Duesing, Strategic Research

W

hat gratitude could you feel for these ruthless gladiators — these brutish, soulless beasts who breathe fire, wield axes, and ram each other until one of them lies dismembered? Combat robots undoubtedly satisfy a deep boyish urge to wreck stuff. But a closer look into the sport of combat robotics reveals something more. The escalating war of robots produces some surprising spoils. As you enter a technological future dominated by satellites, wireless gadgets, and hybrid cars, you may have these evil-natured robots to thank.

Life After BattleBots The sport of combat robotics first entered the public consciousness through the BattleBots show on Comedy Central, which aired five tournaments from 2000 to 2002.

32

SERVO 11.2007

“As far as the sport goes, some robot builders argue whether not being on TV anymore is good or bad,” says Billy Moon, leader of Team Moon Robotics. “On one hand, the show gave the sport a lot of recognition. No matter where we go, inevitably people have seen one of those shows.” The sport received so much attention that it briefly became a piece of pop-culture currency. The trappings of BattleBots — the glitzy graphics and overexcited announcer commentary — also gave the broadcast a veneer of manufactured hype. During the show’s reign, fighting robots were parodied on The Simpsons and the Tonight Show. In the five years since the last

BattleBots aired, much has changed about bot bouts. The events organized under the new national Robotic Fighting League (RFL) are austere and down-to-business. Design and strategy has taken center stage. “I think overall it’s been good to be out of the spotlight,” says Moon. “It’s eliminated the people who just wanted to get on TV. It’s let the sport progress the way it should.” Moon started building robots for himself when he was only 10. Now, at the age of 46, he works at Cisco Systems as a Distinguished Engineer, the highest rank of technical professional. Only a couple of dozen multi-disciplined “Ninja-Class” engineers work for the firm, taking on special projects that require the most out-of-the-box solutions. During his professional career, Moon has created more than 200 new patents, at an average rate of one every six weeks. Moon is also a popular guy in the

Duesing.qxd

10/10/2007

8:48 AM

Page 33

office on “Bring your Kid to Work Day,” when he will bring in many of his team’s remote-controlled creations for a parking lot demonstration. Weekends are spent building and fighting robots with the rest of Team Moon, a small coterie of family members and engineering colleagues. Team Moon began competing in combat robots six years ago, at the height of the BattleBots craze. Its early heavyweight robot, Vladinator, dominated many of the televised tournaments. Now active in the larger and independent RFL, as well as the yearly (untelevised) BattleBots contest, Team Moon operates about a half dozen robots competitively, each ranking near the top of their weight class. Most contests are double elimination tournaments of three-minute bouts, where two robots of the same weight class fight to disable each other. The operator controls include a “tapout” button for when the operator wishes to surrender the match and save its fighter from further damage or humiliation. Most fights end with this forfeit button, where one robot obviously dominates. If the three-minute bell rings, judges award points to the contenders based on aggression, strategy, and damage. Typically, it takes two people per robot to steer the action with radio controls — one to drive the bot around the rink, and the other to fire its weapons. Weapons on super-heavy weight robots (around 340 lbs) are not kidding around anymore. In a three-minute match, the offensive maneuvers — consisting of kinetic thrusts, spinning blades, or bursts of flame — push out up to 200,000 joules of energy, pumping from 2,000 amp, 30 volt reserves of electric power. “You might think a good armor would be 1/4 inch thick 4130 or 4340 steel,” says Moon. “Most weapons now will cut through that like butter.” Since the sport has left the TV spotlight, more responsive engines and more sinister hardware have emerged. The top competing machines are now developed through advanced

The Team Moon robot fleet.

engineering software, digitally simulated, and CNC cut.

The Devil’s Workshop “What’s nice about robotics is that it’s a full system: mechanical, electronic, and artificial intelligence,” Moon says. “You have to know a little bit of everything when you’re building a robot, and that to me is very satisfying.” For robot builders fascinated with performance, strength, power, mechanical motion, and the grating sound of metal-on-metal it is appropriate that the most advanced addition to their workshop is, in a sense, a robot itself — a robot that uses mechanicals, electronics, and programming. The biggest addition to the Moon workroom has been a CNC mill, which cuts metal pieces automatically from the computerized part models. Moon purchased one of the first “personal CNC” machines on the market. The new mill, put out by Tormach, Inc., is able to precision-cut the thick titanium armor, but is smaller and more affordable than the historically huge factory equipment. The “personal” in the trend

of CNC can be likened to the first personal computers, where the technology finally became practical for an individual in cost, size, and performance. And with the advent of easy-to-use CAM (the software which converts CAD files into machine cutting paths), CNC technology is becoming closer in practice to just sending a Word document over to a printer. Of course, in this case, the printer is carving out three-dimensional steel parts. “I’m far from being a machinist myself,” admits Moon. “The Tormach is an excellent example on how easy CNC machining is getting. If we can use it, then anyone can do it. The technology the way it is now, it’s very affordable compared to taking your parts to a machine shop every time.” In contrast to factory-sized CNC mills that bottom out at around $30,000, the PCNC 1100 costs under $7,000. “My older boy actually took in a weeks’ course over the summer to do the CAM programming using a software package called CAMWorks,” Team Moon wooden design models.

SERVO 11.2007

33

Duesing.qxd

10/10/2007

8:49 AM

Page 34

A robot named Goosfraba utilizes a flame-thrower as part of its arsenal. A largersized combat bot can exert as much as 200,000 Joules of energy during a single threeminute bout. Photo courtesy of Sam Kronick.

Team Moon’s ax-wielding Eugene fights the spinning menance called the Shrederator, designed by Team LOGICOM. The matchup creates sparks as the two combat robots contend for the 2006 national heavyweight title of the Robot Fighting League. Photo courtesy of Sam Kronick.

says Moon. “He’s interested enough that he’s actually making a few parts on the Tormach machine, which is an amazing thing to do for a high school kid. My objective for getting him to use tools has surpassed my expectations.”

Robots Making Robots Six years ago, a Team Moon robot

began as a cardboard model, then a wooden one. The physical prototypes were tested and tweaked manually, before the metal parts were finally fabricated. “It took us about a year to design it and about six months to build it, because we had to do so much stuff by hand.” Now, the shop can push out the most modern machines in half the time, thanks to an automated design process that is in many ways more advanced than that of some commercial manufacturers. Robots are now fully designed in SolidWorks, a 3D solid modeler. For his newest creation — called Eugene — Moon used a mechanical simulation software (Cosmos) for various mechanical parts, like stress analysis of the assembly, or repair exercises, which used volumetric data to ensure clearances inside the machine for different sizes of tools. The majority of the robot parts are very complex in construction, having a lot of curves and circles that bend in more than two dimensions. “For our purposes, we would benefit from having a CNC machine where we could CAM these difficult shapes, where the machine would do a lot of the thinking, rather than trying to do it manually. The Tormach was a really good choice for us because it was specifically designed for CNC, whereas a lot of smaller mills are conversions of a manual machine. There’s a lot more you can do from day one with a mill that is set up for CNC.” Despite its small size, the one-ton Tormach mill maintains Billy Moon and his son Will working on Star Hawk.

34

SERVO 11.2007

its cutting power — enough to craft the thick titanium armor — due to the rigidity of its base and table, which are made of cast iron. The mill’s four-axis spindle automatically carves the complicated three-dimensional curves — ones which would be impossible to cut by hand — in a matter of minutes. The high rate of innovation in the sport of fighting bots can be seen firsthand in the evolution of pieces on the mill. Given the ability to make a few iterations, robot parts and assemblies evolve into stronger and more effective devices. In the past, this was impossible. Complex parts needed to be ordered at a local machine shop, which would take a few days or weeks of waiting. Now that Team Moon can cut their own parts in the garage, they can speed up the construction process, while enhancing the design. At-home CNC capabilities give the team the ability to refine the robot design as it goes along. “Even two years ago, there were some parts we had to send to a machine shop. I made a mistake during design about the size of the sprocket for this standard go-cart wheel,” Moon says. “I’d have this little support piece on the inside of the sprocket ring to give a little extra support. In order to get the piece made at a machine shop, I really have to order 10 of them to make it worthwhile, because of the set-up costs. “So I’d order 10, get them back, and the support wasn’t as strong as it could have been,” Moon recalls. “It’s just not cost-effective to go make another one again. I’d just have to live with it, and remember to change it next time we ordered parts. Today, I’d just machine another one. The Tormach mill gives us quick turnaround on rework, which has been invaluable.”

The Spoils of Robotic War Do the rapid innovations seen in robot construction have any uses beyond the arena? Given the robots’ warlike disposition, the first thought that springs to mind might be military or police applications. According to Moon, there has been military interest, particularly in the area of defensive

Duesing.qxd

10/10/2007

8:48 AM

Page 35

Will Moon and his friend Patrick Vanderbee sitting on their bot “The Wall.”

armor. Some in the military have seen the sport’s potential as a training tool for both mechanics and strategic thinking. British Air Force cadets take a course in building robots and fight them on the UK bot circuit, the FRA. The biggest impact of combat robotics may be in the commercial realm. “There are a lot of parts we have designed and we ask manufacturers to build for us, which actually may have a lot of use for people,” says Moon. “For example, electric motors. We are very demanding on our motors. We have the highest packed, highest quality motors money can buy. They have to be super rugged, deliver constant power, and be very lightweight. Five years ago, that was just an odd request. Today, having a high performance electric motor is a very interesting thing if you’re a manufacturer of hybrid cars. What would it take to build a hybrid car? You’d need a lightweight, high-efficiency electric motor that’s pretty rugged.” Because he is an engineer of Ninja status, Moon is fortunate enough to work with suppliers who give him test parts in exchange for feedback. Team Moon often gets prototypes of early technology that is inaccessible to normal consumers. “There have been a lot of manufacturers that we work with closely,” Moon says. “One of them has taken motors that were first developed by combat robots into the wheelchair business. Another vendor was in satellite communications, and needed motors to move the parts on satellites. Now it has a whole line of motors, based on what Eugene. they’ve learned from combat robots.” Batteries are also a big factor in hybrid cars. Any electric car designed for practical use has to contend with limitations of battery life, reliable power delivery, and time it takes to recharge at a stop at the gas station — or rather, your future roadside “power plant.”

“There are a number of industries very dependent on batteries. Some battery makers have given us experimental batteries to test out,” Moon says. “The batteries we need for our robots are just unbelievable. We need batteries that we can completely drain in three minutes. I need to cool them down, and then recharge them within 25 minutes before they go out again for another three-minute drain. We need that level of cycling. Five years ago, it was impossible. I have batteries now that can perform like that.” “If you can do that with your battery technology, then you can build power plants for electric cars; you can build laptops and cell phones that can charge in a couple minutes and then last all day. If the demand is great enough, somebody will build it.” Whatever the future is for bottested technology, the more intangible, but perhaps greater, impact of the sport may be on future generations. Robots have brought fathers and sons together, teaching the youth not just about competition, but how to be mechanically self-reliant. Rather than the passing on the old skills of

Moon and his younger son David bending some titanium. David is using the blow torch and Moon is operating the press.

traditional tools to the next generation, the advanced science of robot war imparts kids with the relevant high-tech skills for later professional or entrepreneurial success: computer modeling, CAM programming, and CNC machining. Now that these building technologies have come down to a personal level of use and affordability, the future is wide open. SV

Team Moon Robotics Team Moon Robotics is one of the world’s top competitors in robotic fighting, participating in several events each year. In the 2006 Robot Fighting League Championship, the team placed five robots in the top three of their respective events. The Moon family lives in Cary, NC. Broken Eugene. Photo courtesy of Sam Kronick.

SERVO 11.2007

35

Eady.qxd

10/9/2007

3:16 PM

Page 36

Do You Want it Now? by Fred Eady

Instant gratification comes in many forms. This month, instant gratification comes in the guise of a simple collection of electronic components designed to provide multiple control input interfaces to a highpowered miniature linear actuator. For a robomagician like you, instant gratification is a Firgelli PQ-CIB controller hooked up to a Firgelli PQ12 linear actuator. The PQ12 Linear Actuator

PHOTO 1. The PQ12 linear actuator is very compact. It measures 36.5 mm at its longest extent, 22 mm at it shortest extent, and is only 22 mm thick. The maximum extended stroke is 20 mm.

The Firgelli PQ-CIB controller is designed exclusively to drive the Firgelli PQ series of miniature linear actuators. In a previous issue of SERVO, you saw just how easy it was to build and code a Firgelli L12 linear actuator driver hardware module from scratch. Doing the same for the PQ12 series of linear actuators is almost effortless, as well. However, if building electronic devices from scratch is not in your mechanically-inclined forte or if you don’t have time to solder and need a complete and proven linear actuator platform in a hurry, the PQ-CIB controller is your best solution. In the linear motion discussion that will follow, we’ll take a look at the PQ-CIB controller hardware and firmware. We’ll also perform a preflight walk-around on the PQ12 miniature linear actuator. In the course of our talks and walks, we’ll outline some PIC-based firmware and some very basic PIC hardware to drive the PQ-CIB controller. This article is all about moving stuff. So, let’s get a linear move on.

36

SERVO 11.2007

The PQ12 linear actuator you see in Photo 1 is formally known as a PQ12f. The “f” in the PQ12’s moniker signifies “high-force.” The high-force PQ12f can generate a force of 15N when the actuator is moving at 7 mm/s. A peak force of 18N can be realized when the actuator is moving a bit slower at 6 mm/s. The PQ12f is more than twice as powerful as its faster first cousin the PQ12s. The “s” here means speedy with the PQ12’s linear actuator being able to extend and retract its actuator at 27 mm/s. That’s over twice the maximum stroke speed of 12 mm/s provided by the he-man PQ12f. And, by the way, the forces and speeds I’ve just outlined are good for both retraction and extension. The PQ12 linear actuators both interface to their control circuitry by way of a Flex-PCB cable. A 1 mm pitch FFC/FPC connector can be used to access the PQ12f’s Flex-PCB cable edge contacts. However, you may also choose to solder lead wires into the solder holes of the PQ12’s cable. There’s no reason to jeopardize your PQ12 Flex-PCB cable as you can obtain a Hirose FH21 five-pin 1 mm FFC/FPC connector from Digi-Key. Unless you specialize in wire and printed circuit board (PCB) interconnects, the acronyms FFC and FPC are a foreign language you don’t speak. In a nut shell, FFC stands for Flat Flexible Cable. FFC cable consists of thin rectangular copper conductors that are laminated between two layers of polyester insulation. A stiffener is attached to the cable end to give it the strength needed for connector mating

Eady.qxd

10/9/2007

3:17 PM

Page 37

PHOTO 2. The integrated battery pack is designed to be detached from the controller. The PQ-CIB controller comes ready to rock out of the box. All you have to do is add some battery power, attach a PQ12, and move the slider to activate the PQ12’s plunger.

and unmating. Due to the way FFC cables are constructed, they are best suited for straight one-to-one connections. Then, there is FPC, or Flexible Printed Circuit, cable. The idea behind FPC cable is identical to the FFC cabling concept except the conductors of FFC cabling are etched. This makes FFC cabling the choice for custom applications that may also need an odd cabling geometry. These cabling technologies have been around since the 1970s. So, unless you were born just recently, you’ve seen lots of FPC and FFC cables. You just didn’t know what to call them. Unless we only want to completely extend and completely retract the PQ12f’s actuator, we’ll need to have total control over how far to allow the actuator to move and complete control over which way the PQ12f’s actuator will move. The mini-schematic of the PQ12f’s internals shown in Figure 1 tells us that the PQ12f is logically identical to the L12 linear actuator we talked about in the previous issue of SERVO. That’s good in the sense that if you read about the L12, your learning curve will be zero as far as the operation of the PQ12f is concerned. Whether or not you know anything about the Firgelli L12 linear actuator, it’s rather obvious from Figure 1 that the PQ12f’s internal potentiometer is the key to sensing

the position of the PQ12f’s actuator. The PQ12f’s actuator drive motor requires a power supply voltage of +5 FIGURE 1. This is pretty simple VDC and draws a maximum stuff for such an accurate and powerful device. With current of 250 mA. This the help from a very fast PIC makes the PQ12f perfect microcontroller, we will have as a powerful replacement absolutely no problem in for +5 VDC hobby servos in keeping up with the position of the PQ12f’s plunger. applications that require the hobby servo to emulate a linear actuator.

PQ-CIB Controller Overview The PQ-CIB controller was designed for Firgelli by the BCIT Technology Centre. My PQ-CIB controller is shown in Photo 2.

V+

Q6 STD30PF03L

R3 10K

SCREW TERMINALS V+

BATTERY +

Q1 IRLML2402 H1

H3

C6 R4 10K

R7 10K

R5 10K

CON2

1.0uF Q3

D2 Q4 IRLR3714

IRLR3714 H2

H4 +5VDC

V+

BAS70

5 4 3 2 1

MT1 MT2

R6 0.27

D3

C1 .1uF

IC1 VDD

R14 1.0M

4

14

MCLR/Vpp

VSS

RA0/AN0 RA1/AN1 RA2/AN2 RA4/AN3 RA5

13 12 11 3 2

RC0/AN4 RC1/AN5 RC2/AN6 RC3/AN7 RC4 RC5

10 9 8 7 6 5

OUT C5 1.0uF

1N4004

R12 1K

NOTES: 1. D3 IS REVERSE POLARITY PROTECTION DIODE

H1 RC INPUT

MOTOR FEEDBACK 4-20mA INPUT 0-10 VOLTS INPUT H4 H3 H2 +5VDC

R11 10K C3 .1uF

R10 249

10K R8 10K

PIC16F676

INT

EXT

C4 2.2uF 1 2 3

1

C2 1.0uF

1 2 3 4 5

PGC PGD

MCP1700T5002E/TT

IN LED1

ICSP MCLR

+5VDC

IC2

FH21

+5VDC

6 5 4 3 2 1

RC INPUT 4-20mA INPUT 0-10V INPUT

+5VDC

Q2 IRLML2402

GND

Q5 STD30PF03L

1 2 3

R2 10K

BATTERY HEADER

POT SELECT

R9 10K

SCHEMATIC 1. There should not be anything here a robobuilder like you does not understand. This is simple PICology mixed in with basic DC motor control.

SERVO 11.2007

37

Eady.qxd

10/9/2007

3:17 PM

Page 38

At first glance, the four MOSFETs mounted at the top of the PQ-CIB controller PCB stand out. I don’t know about you, but to me they are screaming H-BRIDGE! A look at Schematic 1 shows us that the PQ-CIB controller is, in fact, mostly an H-bridge. Everything else is done with firmware that resides within the Microchip PIC16F676. The PQ-CIB controller firmware accepts 4-20 mA, 0-10 VDC, hobby R/C, and 1 kHz PWM input signals, which all can be used to move the PQ12’s actuator. Photo 3 is a reconnaissance view of the PQ-CIB controller sans battery pack. At the top left edge of Photo 3, you can see the FH21 FFC/FPC cable socket. If you backtrack to Figure 1, you can easily figure out how the linear actuator FH21 cable connector is laid out. Pin 1 of the FH21 FFC/FPC is located at the bottom of the FH21 FFC/FPC cable socket. Note that the FH21 FFC/FPC cable socket pins out all of the odd numbered signals on the left and the even numbered signals on the right. You can check this against the pin layout shown in Figure 1. The PQ-CIB controller’s PIC16F676 comes preprogrammed with the necessary stuff to accept all of the signals I ran down earlier. However, you’re reading this because you and I are two peas in a pod. You’ll want to program the PQ-CIB controller with your own driver firmware eventually. So, a five-pin PIC programming interface is positioned directly below the FH21 FFC/FPC cable socket. All of the various inputs — which are rendered mutually exclusive by the firmware — are handled at the screw terminals that lie directly below the five-pin PIC programming interface. The PQ-CIB controller firmware is continually scanning the inputs and locks in on the first valid input type it encounters. Once the PIC16F676’s linear actuator controller firmware gets a valid input, it reverts to using that input type exclusively until the PQ-CIB controller is reset.

The PQ-CIB Controller Hardware Schematic 1 tells the tale. The PQ-CIB controller is an H-bridge under the control of a PIC16F676. The simple layout, a socketed PIC16F676, and use of large 1206 SMT components make the PQ-CIB controller easy to repair, just in case you happen to accidentally cause the PQ-CIB controller to release some magic smoke. The use of firmware to replace hardware is evident in Photo 3. We all know the pitfalls of driving H-bridge configurations directly. You don’t want to turn on the wrong MOSFETs when driving a motor. Normally, some sort of gating logic would be placed between the PIC H-bridge outputs and the MOSFET gates to prevent the inadvertent activation of the incorrect pair of MOSFETs. However, the folks at Firgelli and BCIT put together some firmware to safely drive the PQ12 linear actuators without the need for the extra PIC-to-MOSFET gate logic. The portion of the PQ-CIB controller you see in Photo 3 is designed to be separated from the battery pack for use in a

38

SERVO 11.2007

PHOTO 3. The PQ-CIB controller is perforated at the battery pack edge to allow you to separate the battery pack and the main controller electronics. The idea is to allow you to use the PQ-CIB controller as a linear actuator controller out of the box and into your project.

real-world linear actuator project. Once the battery pack is jettisoned, the remaining PQ-CIB controller electronics can be powered from an external power source. To obtain external power, we simply move the jumper from BATTERY to EXT PWR. You can see this jumper block clearly at the bottom of Photo 3. The demonstration potentiometer is also discarded and disconnected when the battery pack is separated from the PQ-CIB controller electronics. With the demo slider absent, we must move the POT/EXT jumper to EXT to utilize the 0-10 VDC input. That’s all we need to say about the PQ-CIB controller hardware. If you’re a robonewbie and H-bridge sounds like something from a Latin text book, check out the H-bridge article that Peter Best did in the July ‘06 issue of SERVO. Let’s take a scenic tour the PQ-CIB Controller firmware.

The PQ-CIB Controller Firmware The PQ-CIB controller firmware’s main program loop polls the three inputs (R/C, 4-20 mA, 0-10V/PWM) and determines the input type. Once the input type is determined, the firmware then vectors to the input’s service function and loops there, processing the input until a reset occurs. While servicing the selected input, the PQ-CIB controller firmware also monitors the linear actuator for stalls. As you might have already deduced, the PQ-CIB controller firmware uses the PIC16F676’s timers and analog-to-digital (A-to-D) converter heavily. From the “flavor” of the source code, I would say the PQ-CIB controller firmware was written using the CCS C compiler. Another good clue that leads me to believe that CCS C is the compiler of choice lies in the opening declaration code: ********************************************************* #include <16F676.h> #include #include <math.h> #use delay(clock=4000000) // 4 MHz system clock *********************************************************

The #use delay statement is a CCS-specific clock definition compiler command. Since there’s no external crystal or oscillator on the PQ-CIB controller PCB, the only logical conclusion is that the PIC16F676 is running on its internal precision 4 MHz oscillator. That gives us an instruction cycle time of 1 µs. The R/C input is the only input that uses the PIC’s interrupt on change feature coupled with an interrupt handler. If you’re not familiar with how R/C hobby servos work, here’s some quick schooling. An R/C transmitter emits a chain of pulses with each pulse aimed at a specific R/C servo. The R/C receiver decodes these pulses and routes them to the target servo. Typically, an R/C hobby servo actuator will operate

Eady.qxd

10/9/2007

3:18 PM

Page 39

over a range of 180°. A pulse width of 1.5 ms will cause the servo to center its actuator at 90°. To rotate the servo actuator to 0°, a 1.25 ms pulse is transmitted. Sending a 1.75 ms pulse will cause the servo to rotate its actuator to the 180° position. Thus, a hobby servo system operates in response to the width of the pulse coming into the servo. The pulse width measurement is defined when the pulse is logically high and the PQ-CIB controller firmware takes advantage of that fact. The PQ-CIB controller firmware will begin a pulse width measurement when it detects a low to high transition on the R/C input. When the pulse transitions from high to low, the value in TIMER0 — which is also called RTCC in the source code — is read and loaded into the variable RC_In_Voltage. The manipulation of TIMER0, or RTCC, is triggered by the PIC’s interrupt on change mechanism. Here’s the code that makes up the RC input interrupt handler:

Prescaling the TIMER0 clock by 4 means that TIMER0 will increment every four instruction cycles, or once every 4 µs. Once everything I/O is set up and the A-to-D converter and TIMER0 mechanisms are primed, the PQ-CIB controller firmware begins its input scanning. The first order of the input scanning business is to turn off the linear actuator motor and hide for a second or so to avoid picking up some noise:

********************************************************* #INT_RA void isr() { if(input(PIN_A5)==1) { set_rtcc(0xF0); //preload timer }

The scanning firmware is contained within an endless loop. The A-to-D converter is called upon repeatedly to read the 0-10V/PWM and 4-20 mA loop analog inputs. If 16 A-to-D converter ticks are collected from an analog input in a sample period, there’s voltage on that particular input. Here’s the code snippet form the ScanInputs function that hunts down incoming voltage or current:

if(input(PIN_A5)==0) { //read the timer and load into a variable RC_In_Voltage = get_rtcc(); } } *********************************************************

Every PIC program requires that the programmer set up the PIC’s I/O structure. The firmware driver for the PQ-CIB controller is no exception. Note that A-to-D definitions and TIMER0 prescale definitions are included within the defacto I/O initialization code: //******************************************************* //* Main program - sets ports and registers //******************************************************* void main( void ) { OUTPUT_A(0x00); // clear port A SET_TRIS_A(0x2B); // A5, A3,A1,A0 = Input // A4,A2 = Output OUTPUT_C(0x00); SET_TRIS_C(0x07);

// clear port C // C5:C3 = Output, // C2:C0 = Input

SETUP_ADC(ADC_CLOCK_INTERNAL); SETUP_ADC_PORTS(sAN4|sAN5|sAN6); // AN4 FB, // AN5 4-20ma, AN6 0-10V //internal clock used with a prescaler of 4 SETUP_COUNTERS( RTCC_INTERNAL, RTCC_DIV_4 ); // enable GLOBAL interrupt ENABLE_INTERRUPTS(GLOBAL); // function to determine source type ScanInputs(); } *********************************************************

********************************************************* void ScanInputs( void ) { unsigned long In_Voltage = 0, In_Current = 0; unsigned long waiting = 0; MotorOff();

// Motor off and delay to avoid // false signal on start up

delay_us(500000); delay_us(500000); *********************************************************

********************************************************* do{ delay_us(500); waiting=0; In_Voltage = GetADCResult( InputV ); In_Current = GetADCResult( InputC ); //if there is a voltage or PWM signal detected if ( In_Voltage > 0x10 ) { V_PWM_I_SourceAp(InputV); } //if there is a current signal detected else if (In_Current > 0x10 ) { V_PWM_I_SourceAp(InputC); } *********************************************************

The ScanInputs function will invoke the V_PWM_I_SourceAp function if either the voltage or current analog inputs produced a voltage. The argument of the V_PWM_I_SourceAp function call is used to determine which analog input to use as a basis for moving the linear actuator motor. The R/C input not only has its own interrupt handler, it also has a special function that is called when voltage is detected on the R/C input: ********************************************************* else { do { waiting++; if(input(PIN_A5)==1) { RCSourceAp(); } }while (waiting<500); } }while(true); } *********************************************************

SERVO 11.2007

39

Eady.qxd

10/9/2007

3:18 PM

Page 40

The V_PWM_I_SourceAp function works on the premise that if the motor feedback voltage is greater than the input voltage, the actuator will retract. On the other hand, if the motor feedback voltage is less than the input voltage, the actuator will extend. Obviously, if the feedback and input voltages are within a preset hysteresis window, the motor will not move. Like the ScanInputs function, the V_PWM_I_SourceAp function is a never ending do-while loop. The linear actuator’s feedback voltage from its internal potentiometer wiper is continually read and compared against the voltage of the selected voltage control input: ********************************************************* void V_PWM_I_SourceAp( unsigned int Input ) { unsigned long In_Voltage = 0, FB_Voltage = 0; unsigned long Previous_FBv = 0; unsigned int Direction = 0, FB_count=0; unsigned int OK_to_ext = 0x01, OK_to_ret = 0x01; do { FB_Voltage = GetADCResult( FBVolts ); do { //store the previous FBv for stall condition Previous_FBv = FB_Voltage; In_Voltage = GetADCResult( Input ); FB_Voltage = GetADCResult( FBVolts ); if((abs(In_Voltage - FB_Voltage)) < hysteresis) { MotorOff(); Direction = Stopped; } *********************************************************

Note that a copy of the linear actuator feedback voltage is stored in a keeper variable every time that the linear actuator feedback voltage is read. The keeper variable is compared to the next feedback voltage reading to check for a stalled actuator condition. Basically, if the new feedback voltage is equal to the keeper variable value, the actuator motor must be stalled. Here are the firgelli.h definitions for

the A-to-D inputs used in the V_PWM_I_SourceAp function: ********************************************************* #define InputV 6 #define InputC 5 #define FBVolts 4 *********************************************************

And, if you’re wondering about the hysteresis value, it is also defined in the firgelli.h file: ********************************************************* #define hysteresis 0x03 //(FYI 0x10 = approx. 0.175V) *********************************************************

All of the inputs are tested against the linear actuator’s feedback voltage using an algorithm that looks like this: ********************************************************* if((abs(In_Voltage - FB_Voltage)) < hysteresis) { MotorOff(); } else if((FB_Voltage > In_Voltage) && (OK_to_ret== Yes)) { OK_to_ext=Yes; MotorRetract(); delay_us(10); } else if((In_Voltage > FB_Voltage) && (OK_to_ext== Yes)) { OK_to_ret=Yes; MotorExtend(); delay_us(30); } *********************************************************

The OK_to_ret==Yes statement is related to the stall sensing code. Note that as the contents of the function are executed, the OK_to_ret variable is loaded with the Yes value, which is also defined as 0x01 in firgelli.h. If a stall condition is sensed, the stall code will equate the OK_to_ret variable to No, which is defined as 0x00 in firgelli.h. The stall code is executed within every input function: SCREENSHOT 1. Play with the values of CCPR1L and bits 4:5 of CCP1CON to change the duty cycle. Changing the value of PR2 will alter the period of the PWM signal. Changing the period also changes the frequency as the frequency is equal to 1/period.

Sources Saelig — www.saelig.com Cleverscope Firgelli — www.firgelli.com PQ12; PQ-CIB Controller; L12 Custom Computer Services, Inc. www.ccsinfo.com CCS C Compiler

40

SERVO 11.2007

Eady.qxd

10/9/2007

3:18 PM

Page 41

********************************************************* if((Direction==Extend)||(Direction==Retract)) { if((abs(Previous_FBv - FB_Voltage))<0x01) { FB_count++; //count threshold indicates a stall condition if(FB_count>0xA0) { FB_count=0x00; MotorOff(); if(Direction==Retract) { OK_to_ret=No; } else if(Direction==Extend) { OK_to_ext=No; } Direction=stopped; } } else { FB_count=0x00; } *********************************************************

Recall that earlier I pointed out the use of firmware by the PQ-CIB controller to replace hardware. Well, here are the core functions that are used to command the H-bridge: ********************************************************* void MotorOff(void) { OUTPUT_LOW( PIN_C3 ); OUTPUT_LOW( PIN_C4 ); OUTPUT_LOW( PIN_C5 ); OUTPUT_LOW( PIN_A4 ); delay_us(10); } void MotorRetract(void) { MotorOff(); OUTPUT_HIGH( PIN_C3 ); OUTPUT_LOW( PIN_C4 ); OUTPUT_LOW( PIN_C5 ); OUTPUT_HIGH( PIN_A4 ); } void MotorExtend(void) { MotorOff(); OUTPUT_LOW( PIN_C3); OUTPUT_HIGH( PIN_C4); OUTPUT_HIGH( PIN_C5); OUTPUT_LOW( PIN_A4); } **************************************

I’ll supply the entire PQ-CIB controller firmware package to you via the SERVO website at www.servomagazine.com.

Extend and Retract You have everything you need to move the actuator of a PQ12, almost. Here’s some PIC code you can play with to generate the kHz PWM signal for the PQ-CIB controller: ********************************************************* TMR2ON = 0; CCP1CON = 0x0F; CCPR1L = 0x4F; T2CON = 0x02; //prescale=16 PR2 = 0x9C; TMR2ON = 1; *********************************************************

I used the code above with a PIC18F2680 running at 10 MHz to generate the Cleverscope shot you see in Screenshot 1. Now, change your PIC clock to 4 MHz and run this code: ********************************************************* TMR2ON = 0; CCP1CON = 0x0F; CCPR1L = 0x5F; T2CON = 0x02; //prescale=16 PR2 = 0xFF; TMR2ON = 1; *********************************************************

As you can see in Screenshot 2, the update rate is a bit faster than 50 Hz, but this 250 Hz, 1.5 mS PWM pulse train will center an R/C servo and the actuator of your PQ12. Fiddle with the CCPR1L value and bits 4:5 of CCP1CON to change the duty cycle of the waveforms I’ve shown you. The CCP1CON bits are the least significant bits of the PIC18F2680’s 10-bit PWM resolution range. If things go as they should for you, the PWM waveform will be present on RC2. We can stop now because you do have everything you need to put a Firgelli PQ12 linear actuator to work in your next robotic project. Feel free to email me at [email protected] if you have any questions or comments. SV SCREENSHOT 2. R/C servos want to see a pulse every 20 mS, which equates to a 50 Hz PWM signal. This shot is updating every 4 mS. I tied this signal to my PQ-CIB controller and my PQ12 jumped to center stroke.

The H-bridge MOSFET components would have to fail for any illegal MOSFET turn-ons to occur. You can take the MOSFET commands and map out which MOSFETs are on for retract and extend operations. Since MotorOff turns all of the MOSFETs off, we can logically conclude that a logical high will turn on a MOSFET in the H-bridge. Okay, we’ve taken a look at all of the building block code that makes up the PQ-CIB controller firmware package. This walk-through was by no means a line-byline journey. So, with thanks to Firgelli and the folks at the BCIT Technology Centre, SERVO 11.2007

41

Doerr.qxd

10/10/2007

8:24 AM

Page 42

Breathing new life into an old robot ... by Robert Doerr

I

really enjoy working on the older personal robots from the ‘80s. It’s fun to keep them running and I am always looking at ways to enhance them while keeping the original robot intact. Recently, a friend asked me to fix up his old Androbot TOPO robot. This revived my interest in a project that I have been thinking about for quite some time. I wanted an easy way to use my own TOPO robot without setting up an entire Apple II+ system to support it. Another reason is that the original RF transmitter modules required to run the robot usually get lost. Without the transmitter, TOPO can’t do anything. For the most part, this has

42

SERVO 11.2007

caused a lot of TOPO robots to become nice display pieces instead of working bots.

Background Information on TOPO Back in the early ‘80s, Androbot manufactured and sold TOPO robots. There were a few different versions, but for the most part there were two distinct varieties. The early ones (TOPO I) are RF based control (open loop), which allows just the remote control movements of the main drive motors. These have a single white button on the head, silver trim, and red/green LEDs near the wheels

and on the body, which show the direction of movement. The later TOPO II/III robots switched to a bi-directional I/R link (closed loop), had some smarts onboard, and could talk. These had a red button on the head where the I/R transceiver was located, black trim, a black cutout for a speaker, and red LEDs near the wheels. Some of the other LEDs were eliminated. The TOPO III is basically the same as the TOPO II, but eliminated the rest of the LEDs and had a slightly different body which replaced the fold down arms with a mount that could accept optional drink and food trays.

Doerr.qxd

10/10/2007

8:24 AM

Page 43

This article focuses on the original TOPO I robot. All references to TOPO in the rest of the article apply only to this original RF based version.

Original Control System One nice thing about the way the RF control is on the TOPO is that it seems to be patterned after a standard 27 MHz radio control that would be used for model planes. The main board has an LM1872 and logic ICs to receive and decode all the signals. It converts them to ones similar to what a standard R/C receiver would produce. These then go to the rest of the circuitry onboard which acts like a pair of R/C to H-bridge controllers. It also has some drivers for the lights, but that’s about it. The whole thing acts just like a standard R/C receiver with a pair of R/C to H-bridge controllers all on a single board. Standard R/C to H-bridge controllers are common as they are used extensively in combat robots (I used them in my Battlebots) and can be driven from microcontrollers, like the BASIC Stamp. The standard setup to run the TOPO I is a bit involved. It requires a custom Apple I/O card with an AM9513 timer chip to simulate the regular R/C signals which then drove an external custom 27 MHz RF transmitter. To complicate matters, the I/O board only seemed to work with an Apple II+ and not the more common Apple IIe system. The driver software was written in machine language with a portion in Basic. To control the robot, you would run a program on the Apple to drive it around with the computer’s joystick, or you could write your own Basic programs and call the driver code to control the movements. It is not the easiest thing in the world to set up and it leaves a bit to be desired in accuracy of the movements! A little known fact is that the receiver in the robot will also accept standard R/C signals provided the transmitting channel is correct and they are in the right format. This is what this article is about.

Breathing New Life Into an Old Robot After some research, it seems that all the TOPO robots used 27.145 MHz as their operating frequency. That crosses over to Channel A4 in the R/C world. TOPO uses Channel 1 as an analog channel for the left wheel and Channel 2 as an analog channel for the right wheel. The other channels are unused. Since I didn’t want to alter the robot at all (in case I want to set up the old Apple II+ again), I needed to perform any required modifications on the transmitter side. We really only need two channels, but most two and three channel radios map these to a single stick and would not do the proper mixing for a dual drive system. Instead, I wanted to try a dual stick radio for tank style steering. It turns out there aren’t too many 27 MHz dual stick radios available, but I did find one that was perfect. It is the Futaba 4VWD-AM four channel radio on Channel A4. If Channel A4 isn’t available, you can order the radio on any 27 MHz frequency along with an A4 crystal set, then change the crystals. Without knowing for sure if this was going to work, I went ahead and purchased one from Tower Hobbies. With shipping, the radio runs around $120 plus a little extra for batteries. Luckily, this whole project turned out to be easier than I thought it would. First, I had to make sure that TOPO was operating the way it should and that it had a fresh battery charge. Then, I set up TOPO on blocks up off his wheels, turned on the Futaba transmitter, and then turned on the robot. The wheels moved slightly. I moved the right stick up and down and the right wheel moved (but in the wrong direction). Moving the right stick left and right made the left wheel go forward and backward. The left stick needed to control Channel 1 instead of Channel 3. This was a fairly easy mod to make.

Screws to remove the joystick.

The Radio Modification With the radio powered off, remove the four screws on the back of the radio and pull off the back cover. They show how to do this in the manual, as you may need to pull off the back cover if you want to adjust the tension on the joysticks. The modification is straightforward. You don’t have to touch the main board or alter the RF section at all. This is important since doing that may make the radio broadcast out of spec and may cause issues with the FCC. What we want to do is just change what potentiometer goes to each channel. In our case, we just want to swap Channel 1 and Channel 3. We do this at the joysticks themselves. Three screws hold each jotstick assembly. There is a fourth screw near the top center, but it’s a fake one so don’t try to take that one out! Unscrew each of the joystick The wires to swap on the joysticks.

SERVO 11.2007

43

Doerr.qxd

10/10/2007

8:25 AM

Page 44

Servo will most likely reversing switches. be backwards

assemblies but do not remove them completely. We just need to lift them up a bit for clearance to unsolder and swap the wires. Once that is done, unsolder the three leads on Channel 1 (right/left on right stick) and Channel 2 (up/down on left stick). Solder wick works well if you don’t have any special de-soldering equipment. There is a plastic housing holding the three wires together for each channel so the leads won’t get mixed up. Just reroute each set of wires to the other joystick and solder them back in place. The spacing is set up so it will only go in one way. Once that is complete, screw the joystick assemblies back in place and re-install the back cover. If you unplugged the power lead to set the back cover off to the side, then make sure to plug that back in before putting the cover back on. Now when you turn on the radio and turn on TOPO, the right stick should make the right wheel move and the left stick should make the left wheel move. The directions

though. This is easily corrected. On the front of the transmitter, there is a set of switches that control the direction of travel for each of the servo channels. Just flip the switches for Channel 1 and Channel 2 to REV. The transmitter should now move the wheels in the proper directions with the correct joysticks. If there is any movement of the wheels when the joysticks are centered, then adjust the trim tabs on each joystick until the wheels stop with the sticks centered. When properly adjusted, all of the LEDs should be off and will light when the wheels move. That’s it! With this, you now have a large remote control robot that you can easily turn on and use at any time without the grief of setting up a whole Apple II+ system to support it. This makes it much easier if you just want to take TOPO somewhere to drive it around. The biggest benefit is for those people that have a TOPO robot and have never been able to use it. Now you can finally see it do something and have some fun with it!

Ideas for Further Enhancements There are some other things that you can do with the robot to easily add more features. This can be done by

using the extra two channels (3 and 4) on the radio to control extra devices. It can make TOPO a bit more interesting than just taking him for a spin. You could use small servos in these ports for other things. Perhaps you may want the head to move. You could use one of these extra channels for that purpose. Another option is to use an R/C switch to control other devices. Some that I’ve used with excellent results are made by Team Delta. Once a certain threshold is reached on the channel, it can turn on a relay or other device. One in particular has two relays on board. If one of these were plugged into Channel 4, then moving the left stick side to side will energize the appropriate relay. This could be connected to a module with canned phrases so when driving the robot around via the remote, it can say a few things. Get creative and have some fun!

Some Plug-in Enhancements There are a couple ways to proceed. First, you can just use the Futaba receiver that comes with the radio. The first two channels (1 and 2) will remain empty while Channels 3 and 4 are open to use. You don’t even have to touch the original TOPO mainboard if you use a separate battery pack for the receiver. With this setup, the original receiver on the robot handles the drive and the Futaba receiver handles the extra stuff. Another option which I chose to use leverages some of the untapped Upgrade board only.

Upgraded board plugged into the TOPO main board.

44

SERVO 11.2007

Doerr.qxd

10/10/2007

8:26 AM

Page 45

TOPO I electronics.

potential of the original robot. The LM1872 receiver chip on the main board normally decodes four channels (two analog and two digital) but with a couple extra ICs, it can actually decode eight channels. This is the way that it was done on the TOPO main board. They just happen to be routed to an unused eight pin expansion connector at J2 on the main board. The connector J2 (starting at pin 1) provides +24V, GND, CH5, CH6, CH7, CH8, CH3, and CH4. Since Channels 1 and 2 are dedicated to the main drive motors, those are not routed to the expansion connector. These extra channels respond to the Futaba radio and are available. Since we’re using a four channel radio, only Channels 3 and 4 work. I did look at all the signals with the scope and it appears that the Futaba radio actually transmits six channels. It may be possible to modify the radio (depending upon its chipset) and add a couple extra switches or controls to leverage the extra channels but I’ll leave that up to you to experiment with. Finding all these signals on J2 was fantastic. It makes it easy to expand the robot without the need for another receiver in the robot. I just made a small adapter that plugged into J2. The 24V from pin 1 and the ground on pin 2 goes to a couple of voltage regulators to generate +5V

and +12V power for servos, solid-state switches, and other devices. There are also some filtering caps to help condition the power. The signals for Channels 3 and 4 on pins 7 and 8 can be routed to a three-pin header to connect to a standard servo or solid-state switch module. To test it, I connected a pair of regular Futaba R/C servos to these two extra channels. Sure enough, moving the joysticks side to side on Channels 3 and 4 controlled the servos perfectly. One really cool idea is to try something like the SpeakJet (SpeakGin) chip in its R/C mode. That way, just by moving the joysticks from side to side the robot can say a few things while moving. For those that may want to do a little more with TOPO, there is another The original Apple II + I/O card and transmitter.

J2 expansion connector.

gem that I found while going over the main board. It is the odd jumper installed on connector J3. This is where the decoded signals for Channels 1 and 2 connect to the drive system. By removing that jumper, it is possible to isolate the receiver section from the drive system and feed your own signals to the drive electronics. You need to feed the R/C signal into pin 7 for the right motor and in pin 8 for the left motor. The ground signal can come from J2, as well as any required power. Since these are standard servo style signals, a regular receiver on a completely different frequency can be plugged in here. This

References Robot Workshop www.robotworkshop.com Author of this article and service/repairs/upgrades of old robots Robot Gallery www.robotgallery.com Information about the early personal robots including Androbot robots Tower Hobbies www.towerhobbies.com Supplier of Futaba four Channel AM radio Team Delta www.teamdelta.com R/C switches and relay modules Parallax www.parallax.com BASIC Stamp modules

SERVO 11.2007

45

Doerr.qxd

10/10/2007

6:49 PM

Page 46

is great If you want to drive around two robots independently since all the TOPO robots I’ve seen come on the same frequency. A more interesting option is to use a small microcontroller (such as the BASIC Stamp) to send the R/C style signals to the original TOPO H-bridge

circuitry for controlling the robot. This would give TOPO some sort of brain of his own. Going this route, you wouldn’t need any transmitter at all.

Final Thoughts If done carefully, many of these

upgrades and modifications can be done without altering the original robot. That way, you can always go back to the way it was made. I think that is important since these are classic robots and worth being preserved. Just something to keep in mind. SV

C Stamp™ Controller

• LOW COST - High Pin Count • 48-Pin Module with Analog • High Speed, Large Memory • Professional SW Dev. Tools • Easy to Connect Accessories • Free Expert Technical Support Toll-Free: 1 (800) 985-AWIT

www.c-stamp.com

I M S e r v ic e

D e sk to p C N C S o lu tio n s

P .O . B o x 1 4 2 , N e w H u d s o n , M I 4 8 1 6 5

Electronic Parts & Supplies Since 1967

Hobby Engineering Robot Kits, Parts, Tools and Books www.HobbyEngineering.com

46

SERVO 11.2007

P h o n e : 8 8 8 -4 5 1 -1 6 7 0 o r : 2 4 8 -4 8 6 -3 6 0 0

i m s r v .c o m For the finest in robots, parts, and services, go to www.servomagazine.com and click on Robo-Links.

Miller2.qxd

10/9/2007

10:09 AM

Page 47

Building an

ANDROID ARM

by Mark Miller

Part 2

Last time, the main drive systems were built in detail. Now we are ready to put the arm together. Mounting the Wrist Motor The SMS-40 stepper motor uses a square flange with four mounting holes. These holes are for 6-32 size machine screws. You can mount the motor using a small 1/2 inch x 1/4 inch piece of aluminum stock and drilling and tapping holes in it that match the motor hole pattern along one edge (two holes). This will rigidly secure the motor to the stock and, in turn, can be drilled and attached to the forearm plate. Carefully align the gears on the motor and wrist, and attach. You can use 1/4 inch long screws for both attachments. Remember to properly mesh these gears using 110 lb paper strip to assure a perfect fit. Test the rotation by turning the motor gear with your fingers, mak-

ing sure there is no binding or interference. Figure 1 shows the proper mounting relationship on the assembly. You will also note that the through shaft into the wrist is not obstructed by the motor, maintaining access to the end effector. Other materials/methods could also be employed to mount the wrist motor. Feel free to experiment.

The Elbow Joint Figure 2 shows the resin joint system that I use in all my designs, however, you may design and fabricate your own from a variety of materials. Keep in mind the material should not be of a flexible nature and should utilize sufficient surface area to make the joint mechanically “tight.” Between each joint half shown is a 1-1/4 inch round

fender washer that provides a wear surface and smooth precise motion throughout the range of joint movement. These joint halves are secured to the forearm metal plate using two 8/32 x 1/2 inch machine screws, drilled and tapped into the resin. The lower bicep joint half is fastened in the same manner. You will note in the photo, the bicep plate has been sanded to fit the joint half, making a shapely finished appearance. Figure 3 shows the completed joint ready for final assembly. The center 1/4-20 bolt holding the two joint halves is clearance-holed in the forearm portion and tapped into the bicep portion. This manner of construction allows the joint to be preloaded

FIGURE 1

FIGURE 2

SERVO 11.2007

47

Miller2.qxd

10/9/2007

10:10 AM

Page 48

FIGURE 3

FIGURE 4

(tensioned) to remove any play between the parts, yet smoothly transition in a full range of motion without drag.

The Shoulder Joint To replicate the shoulder’s motion, a three axis system was devised, consisting of three distinct components. Figure 4 shows the right angle nature of the first joint component from the bicep. This is a 90 degree angle from the inside of the bicep which will face toward the body. This joint will allow the arm to move in and away from the body (as in raising your elbows up even with your shoulders). This joint can be manually “locked” into position or motorized with a pusharm system later, if desired. Figure 5 shows the mating joint surface added FIGURE 6

48

SERVO 11.2007

to allow rotation of the arm forward or back. The joint is assembled as all others with a single 1/4-20 bolt and fender washer in between. Figure 6 shows the final assembly from the outside, while Figure 7 is the backside. It is now ready to be mounted to a test stand for further work.

Getting Things Moving You will find it considerably easier to do the next operations if the arm is fastend temporarily to a test stand. The one shown in the photos was made from wood scraps. The arm is attached at the new shoulder joint by drilling and tapping into the shoulder hub and securing the joint to the stand, FIGURE 7

FIGURE 5

supporting the arm at a right angle, as shown in Figure 8. To move the arm at the elbow, a circular disc of aluminum was used with an OD of 1-3/4 inches. It is 3/4 inches thick and secured to the bicep out shaft from the gearing with two set screws and right angles to each other for maximum holding force. Figure 9 shows the installed disc. You could also use a pulley, instead. For maximum joint movement, a machine screw has been tapped near the outer edge of the disc, from which a drag arm will be mounted extending down to the forearm. This, in effect, will be a simplified crankshaft-type drive system. You can select the best distance from the forearm joint you wish to mount the drag arm with the following in mind: The closer you place the drag FIGURE 8

Miller2.qxd

10/9/2007

10:13 AM

Page 49

FIGURE 10

FIGURE 11

FIGURE 9

arm to the elbow joint, the more range of motion you will receive. However, this is at the cost of overall torque available at the wrist. The photo depicts the drag arm (1/2 inch wide, .062” thick aluminum stock) located at 2-1/2 inches from the joint center, allowing a two pound load to be carried at the wrist. You can experiment for best results for your application. Figure 10 shows the drag arm link connected to the forearm with a short stub of 1/2 inch aluminum drilled and tapped at both ends for 8-32 machine screws. I used a piece of brass tubing under the drag arm screw so the threads were not riding against the metal surface itself. After attaching, run the motor to test for binding or other problems. You will note the whole joint is very rigid and precise with excellent repeatability. There should be no slop in the assembly. The bicep disc also makes an excellent place to mount an encoder which will allow you to keep track of the arm’s position. You may note — due to the offset crank-style design — the arm can be set in continuous motion in either direction without the need for stops or limit switches.

A Simple End Effector A simple claw-type gripper can be

fabricated from 1/4 inch Lexan material as shown in Figure 11. This one was made so one side of the gripper is fixed (upper in photo) and the lower portion (with a small protruding leg near the joint) opens. These have been drilled and tapped into the custom wrist hub using 6-32 screws. To make the movable side open, a small piece of piano wire was run through the wrist joint tubing and pushes against the protruding leg on the claw. I drilled about 1/4 inch into this leg so the wire would not jump out during wrist rotation. As a simple actuator, the tubular pusher solenoid mounted in the foream opens the claw to its maximum (about 1-1/2 inches) when powered. The connection between the solenoid plunger and the piano wire is a small leftover Lexan cutoff epoxied to both. The wrist can operate while the claw opens and closes. This is detailed in Figure 12. A return spring is mounted across the flat part of the claw to keep it in tension while unpowered. The spring has the maximum amount of tension that the solenoid is capable of overcoming. You can experiment and easily determine this tension for your solenoid. Figure 13 shows the tensioning return spring

FIGURE 12

across the fingers. This is a very simple gripper made from scrap. You can purchase a conventional servo-type gripper from several sources that may be more suited for your purpose. The power leads from such a gripper can be fed through the wrist tubing, as well. Driver circuitry for the wrist and bicep motors can be mounted in any leftover space on the arm. There is plenty of open flat area on the aluminum plates to mount some standoffs and the circuitry.

Exterior Covering The covers shown in Figure 14 were made from a fiberglass layup over a cardboard tube. Cutouts were made FIGURE 13

SERVO 11.2007

49

Miller2.qxd

10/9/2007

10:13 AM

Page 50

FIGURE 14

for clearance and then painted. Other possibilities include styrene sheet or other thin bendable plastics, wood, or even the cardboard tube itself (which I thought about after using it as a mold). Those heavy mailer tubes can be cut easily, sprayed with polyurethane, then sanded and fished, as desired. Use common circuit board standoffs to hold the shield in place from the aluminum plate. This also adds rigidity to the arm’s overall structure.

Shoulder Rotation Motorization Adding this motor system allows the arm to move front to back of the body. In combination with the elbow joint, it allows for the arm to reach out

to objects and pick them up or re-place them. The basic system of motor gear combination is again repeated for the bicep/elbow joint. Since this joint is supporting not only the load of the arm itself plus any payload, the output shaft should be increased to at least 1/4 inch shafting. Everything else is exactly the same. If you need more lifting power, additional reduction can be made accordingly. This new gear/motor reduction is mounted to the third joint component of the shoulder. You will need to remove the arm from its building stand to install this feature. The 1/4-20 tapped hole previously holding the arm to the stand can be redrilled out to a full 1/4 inch to allow the shaft to slide into the shoulder

PARTS LIST ITEM GEARS • Set 92657 • Set GR-86

SUPPLIER/PART NO.

American Science and Surplus All Electronics (www.allelectronics.com)

MISCELLANEOUS • K & S Metals Available at your local hobby shop • Resin joints, humanoid robotic components, including a complete kit of parts (mechanical structure) to build this arm, email [email protected] STEPPER MOTORS • Elbow and shoulder • Wrist motor

50

SERVO 11.2007

Electronic Goldmine/KP4M4-G14781 (www.goldmine-elec.com) Alltronics/SMS40-2401-A (www.alltronics.com)

joint, which has been secured by cross-drilling a cotter pin through the joint material and the shaft. The pin will also serve as a quick release to remove the arm from your robot. Figure 15 shows the orientation of this new network in relation to the existing arm. The shoulder-to-body mount is now accomplished by a single 1/4-20 bolt tapped into the third joint component, and allows the entire arm FIGURE 15 to rotate front to back of the body to which it is attached. This axis can also be motorized using a drag arm/motor system mounted on the body to this plate, or it can simply be postioned and tightened in place parallel to the body. This motion, for instance, would allow both “hands” to touch each other in front of the body for manipulating or passing an object from one hand to the other.

Armed and Ready I am currently offering the basic mechanical components of the arm, including all the joints, aluminum plates, hardware, etc., for $35 plus shipping F.O.B. Florida. You supply the motors and gears, end effector components, driver electronics, etc. Specify left or right arm when ordering. The supplied components will be largely assembled when they reach you. Drill holes for your specific motor mounting and gears, and go! Contact me at [email protected]. I really don’t run a business, I do mostly research, but have built many of these arm designs and have extras as new, unused surplus that I am willing to part with. You can also just do it yourself from scratch. The key to the success of a tight, smooth working joint is surface and surface area. A combination of materials that smoothly move against each other for low friction, and plenty of surface contact area to journal against is the real milestone. Experiment! SV

Simpson.qxd

10/9/2007

11:29 AM

Page 51

by Michael Simpson

Etek EB-85A The Etek EB-85A module shown in Figure 1 is the most accurate GPS that I have ever tested. It supports up to 32 parallel channels with WAAS enabled. The downside is that WAAS is not enabled by default, so you must send commands to the module in order to activate this feature. Once the module is powered down, it loses these settings. Let’s take a quick look at the feature set of the Etek EB-85A: • • • • •

32 channel receiver Built-in antenna High sensitivity: -158 dBm 1-5 Hz update rate Selectable baud rate from 4800 to 115200 • 9.8’ positional accuracy/8.2’ with WAAS • Hot start: 1 second • Warm start: 33 seconds • Cold start: 36 seconds • 55 mA power consumption to acquire and 30 mA for tracking • 3.3–5 volt operation • Outputs NEMA 0183 • Small foot print: 30 mm x 30 mm x 8.6 mm • Eight-pin interface cable included • Free Mini GPS utility available

GPS PART 2

Currently, there is no development board available for the Etek EB-85A, but the Copernicus evaluation board works perfectly. I recommend the following steps in order to connect the EB-85A to the Copernicus board.

purple, then orange. • STEP 2: Break off the black and red leads as shown in Figure 3. Then break off the white and green leads as shown. Place a piece of black tape on the remaining leads.

• STEP 1: Start by cutting the white connector off the end of the included • STEP 3: Only four leads are needed cable. Strip about 1/8” of the insulation from each of FIGURE 1 the connected wires, then connect the wires to an eight-pin male header as shown in Figure 2. I recommend using heat shrink on each of the leads as shown. Be sure to connect the wires to the header in the following order: red, black, white, green, yellow, blue, FIGURE 2

FIGURE 3

SERVO 11.2007

51

Simpson.qxd

10/9/2007

11:29 AM

Page 52

FIGURE 4

FIGURE 5 FIGURE 6

board between two pieces of plastic shown in Figure 7. This will make the board more durable and easier to test. The Etek EB-85A module defaults to 34800 baud, so keep this in mind later when we test the module. Since the EB-85A can be powered by a 5V source, you may use the alternative interface shown in Figure 8. This is simply a 5V regulator and EZRS232 driver available from Kronos Robotics. Schematic 1 shows the actual hookup.

Testing the ETEK EB-85A

You have a couple of choices available to you in order to test your interface. First, you can download the free Mini GPS software available from the Sparkfun webFIGURE 7 site at www.sparkfun.com/data sheets/GPS/MiniGPS_1.32.zip. to connect the EB-85A to the connector marked Port A. Attach Use the port marked Port B on Copernicus board. These are shown in the EB-85A to the foam tape as shown the evaluation board to interface with Figure 4. Connect the appropriate leads in Figure 6. Make sure the patch the module. to the board as shown in Figure 5. antenna is facing up. This software was specially designed to monitor and control the • STEP 4: Place a small piece of • STEP 5: As outlined in Part 1 of this EB-85A as shown in Figure 9. You double stick foam tape on the DB9 series, I sandwiched the evaluation can also use it to turn on the WAAS feature as shown in Figure 10. You can use the QuickTerm FIGURE 8 and Satellites software included in the project download shown in Figures 11 and 12. Remember, the EB-85 does not power up with the WAAS feature enabled. Both these programs send the commands to enable WAAS. I have also included the ZeusPro source code used to make these applications. The function ‘turnonwaas’ is used to send the commands to turn on the feature. Be sure to download the MTK Packet manual from Sparkfun. This manual lists all the commands used to configure SCHEMATIC 1 the EB-85. FIGURE 9

52

SERVO 11.2007

Simpson.qxd

10/9/2007

11:30 AM

Page 53

Copernicus GPS Module I have used the Copernicus evaluation board so much that it only seems fair that we have a look at the Copernicus GPS module. It was meant to plug into the evaluation board, but with a bit of bending you can plug the board into a standard breadboard. Just keep in mind that you will need to create a 3.3V interface. Let’s take a quick look at the feature set of the Copernicus: • 12 channel receiver • External antenna connector • High sensitivity: -152 dBm • NEMA 0813, TSIP, and TAIP protocols supported • Cold start: 39 seconds • 28.5 mA power consumption • 3.3–5 volt operation • Small foot print: 30 mm x 30 mm x 8.6 mm • Free GPS utility available You may have noticed that the Copernicus module does not support WAAS. It is rumored that a WAAS firmware upgrade will be made available in the near future. • STEP 1: There’s not much to the installation of the Copernicus module. You simply slip the module into the evaluation board as shown in Figure 14.

FIGURE 10

FIGURE 11

e-339488/Copernicus_Monitor_ V1-02-07.exe. Use the port marked Port B on the evaluation board to interface with the module. The software called Trimble GPS Monitor (Figure 16) is quite extensive and comes with its own manual. The monitor allows you to log GPS data and monitor key data points such as satellite signal strength as shown in Figure 17. It also has a real time plotter shown in Figure 18. I have created a special version of the QuikTerm and Satellites programs for the Copernicus module. The

• STEP 2: I recommend you sandwich the board between two pieces of plastic as we did earlier. You will also need to connect an external antenna to the module as shown in Figure 15. The Copernicus module does not come with a built-in antenna.

FIGURE 13

FIGURE 14

Testing the Copernicus Module Again, you have a couple of choices available to you in order to test the Copernicus interface. First, you can download the free Copernicus Monitor software available from the Sparkfun website at http:// trl.trimble.com/dscgi/ds.py/Get/Fil

FIGURE 12

FIGURE 15 SERVO 11.2007

53

Simpson.qxd

10/9/2007

11:31 AM

Page 54

FIGURE 16

Copernicus module does not display some of the NEMA messages we need by default. The two programs included in the download send special commands to the module to enable the needed messages. Again, I have included ZeusPro source code so you can modify and create your own programs.

Holux GPSlip236 Receiver

FIGURE 17

Last month, I briefly mentioned the Holux GPSlip236 shown in Figure 19. Due to its versatility, I feel a more detailed look at this receiver is warranted. The GPSlim236 has several things going for it. Take a look at some of the features: • • • •

FIGURE 18

54

SERVO 11.2007

USB/Serial interface Bluetooth interface TTL Interface SiRF III chip set

FIGURE 19

• Supports both SiRF binary and Nema 0813 formats • Built-in (replaceable) Li-ion battery gives 10 hours of operation • Can be powered by 5V • On/Off switch with LED indicators • Home and auto adapters/chargers included • Compatible with PC, laptop, tablet, PocketPC, Smartphones • WAAS compatibility • Built-in antenna • External antenna connector While the features list shows WAAS compatibility, I have yet to see this feature manifest itself with any application. Even the GPS viewer software that has an option to turn this feature on fails to do so. What does this mean? Forget about WAAS on this device. Even without WAAS, this receiver outperformed most others I tested. The two easiest ways to communicate with the GPSlim236 is via the optional USB cable or a Bluetooth dongle shown in Figure 20. The small USB connector is not a true USB interface. To connect the module to your PC using USB, you need a special proprietary USB cable which is not included with the receiver. The actual output format of the mini USB connector is asynchronous serial at 38400 baud at a TTL level. This makes it perfect for interfacing to a microcontroller, or as an option, you can build an RS-232 interface as shown in Figure 21. Notice how I used SchmartBoard jumpers as interconnects. The antenna connector can be connected to the Sparkfun adapter as shown in Figure 22. This allows you to

FIGURE 20

Simpson.qxd

10/9/2007

11:31 AM

Page 55

connect the same antenna that we used on the Copernicus and other modules tested previously. Just like before, I have included a QuickTerm and Satellites version of the test software along with source code for use on the GPSlim. The receiver comes with some test software on the included CD for both the PocketPC and desktop. So, what is all the fuss about concerning this receiver? Bluetooth. With the Bluetooth interface, you can connect this receiver to Pocket PCs and Smartphones that do not have serial interfaces. If you decide you would like to pick up one of these little gems, just visit the Amazon website at www.amazon.com and do a search on GPSlim236.

What’s Next? I ran out of room this month, but next month we will look at creating a data logger that all the modules can use. A data logger is extremely important as it will allow us to collect data that we can use to help us test various aspects of our project without having to resort to testing in the field.

Parts List The following is a breakdown of sources for all the components needed for Parts 2 and 3 of this project.

SMA to MMCX adapter cable www.sparkfun.com/commerce/ product_info.php?products_id=285

SPARK FUN ELECTRONICS

KRMICROS

Etek EB-85A GPS Module www.sparkfun.com/commerce/ product_info.php?products_id=8266

ZeusPro www.krmicros.com/Development/ ZeusPro/ZeusPro.htm

Copernicus Module www.sparkfun.com/commerce/ product_info.php?products_id=8146

KRONOS ROBOTICS

Copernicus Evaluation Board www.sparkfun.com/commerce/ product_info.php?products_id=8145 Nine-Pin Serial Cable www.sparkfun.com/commerce/ product_info.php?products_id=65

5V Regulator www.kronosrobotics.com/xcart/ product.php?productid=16304 EZRS232 Board www.kronosrobotics.com/xcart/ product.php?productid=16167 SCHMARTBOARD

6V AC Adapter www.sparkfun.com/commerce/ product_info.php?products_id=737

Jumpers 5” Yellow www.schmartboard.com/index.asp? a=11&id=42

External Antenna with SMA connector www.sparkfun.com/commerce/ product_info.php?products_id=464

Jumpers 3” Red www.schmartboard.com/index.asp? a=11&id=41

We will also start looking at the various microcontroller interfaces for each of the modules. Be sure to check

forupdates and downloads for this article at ww.kronosrobotics.com/ Projects/GPS.shtml. SV

FIGURE 21

FIGURE 22 SERVO 11.2007

55

RoboResources.qxd

10/10/2007

4:53 PM

Page 56

Tune in each month for a heads-up on where to get all of your “robotics resources” for the best prices!

Learning Robotics From the World’s Robotics Labs Y

ou’ve heard the saying, “Imitation is the sincerest form of flattery.” It’s true even for robotics. For even the expert builders, many of the best ideas are cribbed from other people’s designs. Any given robot is 98% the same as all the other robots; it’s the 2% that makes your creation special. Not only is this natural (in any endeavor), it’s to be expected. Progress is based on the many small steps taken by others before us. It’s up to us to incrementally improve upon the ideas, making our own innovative stamps.

If you’re been reading this magazine for any length of time, you’ve discovered all sorts of new ideas you can try out. It’s a great source of inspiration. If you’re looking for still more ideas, here’s yet another source: the university research labs that publish their designs and results on the Internet. Many of the concepts and constructions designs we take for granted today were born in these labs. Over time, the concepts have become part of the general robotics culture. In this column, we’ll review several

Biorobotics is dedicated to the advancement of the field of robotics using insights gained through the study of biological mechanisms.

dozen of these labs. Some of the sources are the official Web pages of the robotics labs at public and private schools and universities; other sources include informal clubs at various schools, as well as government resources. These resources are often rich in details, always free, and often overlooked.

RHIT Robotics Team www.rhitrobotics.org According to the site, the task of RHIT is to design, build, and maintain an aerial vehicle which is capable of interacting with its environment. Annual competition. Sponsored by Rose-Hulman Institute of Technology, in Terre Haute, IN.

Australia Telerobot http://telerobot.mech.uwa.edu. au/Telerobot/index.html Telerobotics from the down under.

BARt-UH www.irt.uni-hannover.de/irt/asr BARt-UH is a bipedal autonomous walking robot designed at the University of Hannover in Germany. Web page is in English.

Biorobotics http://biorobots.cwru.edu Bots at the Biologically Inspired Robotics Lab at Case Western Reserve University. According to the site, the lab “is dedicated to the advancement of the field of robotics using insights gained through the study of biological mechanisms.”

56

SERVO 11.2007

RoboResources.qxd

10/9/2007

11:09 AM

Bruno Jau Robotic Hand http://uirvli.ai.uiuc.edu/tlewis/ pics/hand.html

Page 57

Be sure to check out the archives:

www.cs.cornell.edu/Info/Projec ts/csrvl/csrvl.html

Finally, a robot that will give you a hand.

1996 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots96.html

Past and current projects at Cornell Robotics and Vision Laboratory in Ithaca, NY.

Carnegie Mellon — The Robotics Institute www.ri.cmu.edu/home.html

1997 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots97.html

DEMO — Dynamical & Evolutionary Machine Organization www.demo.cs.brandeis.edu

1998 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots98.html

From the site: “DEMO attacks problems in agent cognition using complex machine organizations that are created from simple components with minimal human design effort.” From Brandeis University in Waltham, MA.

All about the Robotics Institute at Carnegie Mellon University in Pittsburgh, PA.

Carnegie Mellon University: Minerva www.cs.cmu.edu/~minerva/ tech/index.html Minerva is an autonomous tour guide — “we’re walking, we’re walking, we’re stopping ...”

Case Western Reserve University — IGERT http://neuromechanics.cwru.edu About the Neuro-Mechanical Systems program at Case Western Reserve University in Cleveland, OH.

Cognitive Architectures http://ai.eecs.umich.edu/ cogarch2 Online articles compare a variety of current proposed cognitive architectures and a workable structure for classifying and comparing future proposed cognitive architectures. See also:

1999 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots99.html 2000 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots00.html 2001 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots01.html 2002 archive: http://ranier.hq.nasa.gov/telerobotics _page/coolrobots02.html

Cornell Robotics and Vision Laboratory

See also The Golem Project, www.demo.cs.brandeis.edu/golem

Field Robotics Center www.frc.ri.cmu.edu Carnegie Mellon University. Research into sun-synchronous navigation will discover, express, and exhibit the importance of reasoning about sunlight as it pertains to robotic exploration.

Franklin Institute’s Robotics www.fi.edu/qa99/spotlight2 Robotics at the Franklin Institute

“Cool Robot of the Week.”

http://ai.eecs.umich.edu/cogarch0/ http://ai.eecs.umich.edu/cogarch0/ subsump/

Cool Robot of the Week http://robotics.nasa.gov/archive /crw_archive.php and http://ranier.hq.nasa.gov/tele robotics_page/coolrobots.html According to the website, “The honor of being listed as ‘Cool Robot Of The Week’ is bestowed upon those robotics-related websites which portray highly innovative solutions to robotics problems, describe unique approaches to implementing robotics systems, or present exciting interfaces for the dissemination of robotics-related information or promoting robotics technology.” SERVO 11.2007

57

RoboResources.qxd

10/9/2007

11:10 AM

in Philadelphia, PA.

Georgia Tech Intelligent Systems & Robotics http://robotics.gatech.edu The goal of the Intelligent Systems and Robotics group at the College of Computing at Georgia Tech is to understand and design systems which use intelligence to interact with the world, making computer controlled systems more autonomous and ubiquitous.

Georgia Tech Mobile Robot Lab www.cc.gatech.edu/ai/robotlab According to the website, the Mobile Robot Laboratory’s charter is to discover and develop fundamental scientific principles and practices that are applicable to intelligent mobile robot systems.

Hexplorer 2000 http://real.uwaterloo.ca/~robot /www/index.xml The Hexplorer is a six-legged walking robot at the University of Waterloo, located in Ontario, Canada. Construction details and programming overview are provided. See also the main page for the Motion Research Group at http://real.uwaterloo.ca.

Image Science and Machine Vision Group www.ornl.gov/sci/ismv The Image Science and Machine Vision Group is currently involved in three programmatic areas: measurement and controls for industry, biological sciences, and surveillance and security.

Intelligent Systems and Robotics Center (ISRC) www.sandia.gov/isrc/home.html Among other projects, ISRC contemplates robots for warfare and national security. Research includes Surveillance And Reconnaissance Ground Equipment (SARGE), Miniature Autonomous Robotic Vehicle (MARV), Accident Response Mobile Manipulator System (ARMMS), and a Robot That Makes Up Acronyms (RTMUA).

58

SERVO 11.2007

Page 58

Iowa State University Robotics Club http://nukelab1.student.iastate.edu/wiki/Main_Page Iowa State University. Project Cybot is a unique combination of a continuous senior design project and a club open to all students at ISU.

IRIDIA Projects and Activities http://iridia.ulb.ac.be/Projects Artificial intelligence white papers and project summaries. From IRIDIA, the Artificial Intelligence research laboratory of the Universit Libre de Bruxelles.

JPL Rover and Telerobotics www-robotics.jpl.nasa.gov If it walks on another celestial body — like Mars — and is launched by NASA, JPL built it. Here you can read about JPL’s past, present, and future projects. Be sure to check out the Robotic Vehicles Group page.

Laboratory for Perceptual Robotics (LPR) www-robotics.cs.umass.edu University of Massachusetts Perceptual Robotics lab.

LEGO: Distributive Intelligence with Robots www.ceeo.tufts.edu/me94 How they got LEGO Mindstorms robots to work together. Demonstrations include: Whistling Brothers, Travel by Beacon, and Wandering Cyclops.

Machine Intelligence Laboratory www.mil.ufl.edu The goings-on at the Machine Intelligence Laboratory, at the University of Florida in Gainesville, FL.

Massachusetts Institute of Technology http://web.mit.edu This is the main website for the Massachusetts Institute of Technology (MIT) in Cambridge, MA. Links on the main page take you to various labs and research centers at the campus. Spend some time on this one.

MIT: Artificial Muscle Project www.ai.mit.edu/projects/ muscle/muscle.html The Artificial Muscle Project at the MIT Artificial Intelligence Laboratory plays around with linear actuators using a substance known as polymer hydrogel. This material is said to have characteristics similar to human muscle.

MIT: Leg Laboratory www.ai.mit.edu/projects/leglab The Leg Lab is world-renowned for its designs of various single- and multi-pedal robots. Movies are available for many of the designs.

MIT: Logo Foundation http://el.media.mit.edu/logofoundation Informational page about the Logo programming language, originally developed by professors at MIT.

MIT: MindFest www.media.mit.edu/mindfest MindFest is a yearly gathering of LEGO-heads. You can see pictures of past events, and read up on upcoming ones.

Mobile Robots at Loughborough www.lboro.ac.uk/departments/el/robotics There are lots of robots at the Department of Electronic & Electrical Engineering at Loughborough University in England.

Mobility and Robotics Systems http://telerobotics.jpl.nasa.gov/ tasks/tmr Here you’ll find detailed descriptions of the activities of the Mobility and Robotic Systems Section, as well as related robotics efforts around the Jet Propulsion Laboratory. Approximately 100 engineers work on all aspects of robotics for space exploration and related terrestrial applications.

NASA www.nasa.gov The main home page of the National Aeronautics and Space

RoboResources.qxd

10/9/2007

11:10 AM

Page 59

Administration — the outfit in the US that launches the Space Shuttle and the occasional robot.

Robotics and Intelligent Machines Laboratory http://robotics.eecs.berkeley.ed u/wiki/pmwiki/pmwiki.php

NASA JPL: Mars Pathfinder http://mars.jpl.nasa.gov/MPF/ index1.html

Research and activities at the Robotics and Intelligent Machines Laboratory at the University of California at Berkeley.

Informational site about the Mars Pathfinder mission. The mission may be over, but the interest in it is not. See also www.nasa.gov/mission_pages /mer/index.html for current Mars exploration details.

Navy Center for Applied Research in Artificial Intelligence www.nrl.navy.mil/aic/index.php The Navy Center for Applied Research in Artificial Intelligence (NCARAI) has been involved in both basic and applied research in AI since its inception in 1982. NCARAI, part of the Information Technology Division within the Naval Research Laboratory, is engaged in research and development efforts designed to address the application of artificial intelligence technology and techniques to critical Navy and national problems.

Poly-PEDAL Lab http://polypedal.berkeley.edu/ twiki/bin/view/PolyPEDAL/Web Home

Robotics Group at Columbia University www.cs.columbia.edu/robotics Research in robotics — both mobile and stationary — at Columbia University in New York, NY.

Robots at Space and Naval Warfare Systems — SPAWAR www.nosc.mil/robots

Side Collision Warning System for Transit Buses www.ri.cmu.edu/projects/project_ 324.html Just imagine the collision system on a big robot instead of a bus. From Carnegie Mellon University.

Stanford Robotics Laboratory www.robotics.stanford.edu

A look at robotics at SPAWAR (Space and Naval Warfare Systems Center) in San Diego, CA. Most of the robots are for military, urban defense, or other applications where weapons systems — both lethal and non-lethal — are involved. Sandia Intelligent Systems & Robotics Center www.sandia.gov/isrc The Intelligent Systems

Robotics Center (ISRC) is a world leader in creating miniature to macro-sized, teleoperated to autonomous vehicles for military and industrial applications. From environmental clean-up to the battlefield, the ISRC is expert in developing unique intelligent mobile systems.

How robots live at the Robotics Laboratory in Stanford, CA. See also Stanford’s autonomous helicopter work at http://sun-valley.stanford. edu/projects/helicopters/helicop ters.html.

Talking Heads www.haskins.yale.edu/featured/heads/heads.html and

This website provides an overview of the rapidly growing international

Informational page about the Logo programming language, originally developed by professors at MIT.

The Poly-PEDAL Lab studies motion in animals and insects. The walk (gait) and balance studies often help in designing legged robots.

Polypod http://robotics.stanford.edu/ users/mark/polypod.html Polypod is a bi-unit modular robot.This page presents work done in 1993 and 1994. Work on the next generation, called “PolyBot” started mid 1998 at Xerox PARC as part of the modular robotics project under the smart matter theme.

Robotics and Computer Vision Laboratory www-cvr.ai.uiuc.edu The way they see things as the Robotics and Computer Vision Laboratory at the University of Illinois in Urbana, IL. SERVO 11.2007

59

RoboResources.qxd

10/9/2007

2:02 PM

Page 60

effort to create talking heads (physiological/computational/cognitive models of audio-visual speech), the historical antecedents of this effort, and related work. Links are provided (where possible) to the sites of many researchers and commercial entities working in this diverse and exciting area.

Programming Laboratory in Carnegie Mellon’s Robotics Institute.

Tarry Walking Machines www.tarry.de/index_us.html

Overview of the Machine Vision Unit at the University of Edinburgh (that would be in Scotland).

According to the page, the Tarry walking machines which were developed and built by the Department of Engineering Mechanics at the University of Duisburg. Good detailed look at hexapod designs. Recommended reading.

Toy Robots Initiative www.cs.cmu.edu/~illah/EDU TOY The Toy Robot Initiative aims to commercialize robotics technologies in education, toys, entertainment, and art. Operated out of the Mobile Robot

60

SERVO 11.2007

Union College Robotics Club www.vu.union.edu/~robot Schenectady, New York.

University of Edinburgh AI Machine Vision Unit www.ipab.inf.ed.ac.uk/mvu

University of Michigan Artificial Intelligence Laboratory http://ai.eecs.umich.edu People and projects at the AI lab at the University of Michigan in Ann Arbor.

University of Michigan Mobile Robotics Lab www.engin.umich.edu/ research/mrl Among

other

things,

special

research goes on here in the fields of robot navigation. Be sure to read the details of the Mobile Robot Positioning and Obstacle Avoidance research. The book Where am I? (in print, CD-ROM, and electronic download) published by the university’s Johann Borenstein, is a classic and is required reading in many mechatronics courses. See also Dr. Borenstein’s home page where he provides links to many more online robotics resources at www-personal. engin.umich.edu/~johannb.

University of New Hampshire Robotics Lab www.ece.unh.edu/robots/rbt_ home.htm The research emphasis of the Robotics Laboratory in the Department of Electrical and Computer Engineering at the University of New Hampshire is the application of fast associative memories and other neural network learning techniques (such as CMAC neural networks) to problems in

RoboResources.qxd

10/10/2007

1:15 PM

control, pattern recognition, and signal processing. University of Reading Department of Cybernetics www.cirg.reading.ac.uk/home.htm The Cybernetic Intelligence Research Group studies intelligence and its real-life applications.

University of Southern California Robotics Research Laboratory www-robotics.usc.edu Robotics research at USC spans a large number of labs and projects. These include: • USC Robotics Research Lab • Interaction Lab (control and learning in multi-robot and humanoid systems)

Page 61

University. Check out past and present projects, including the stairclimbing robot and the omnidirectional wheel designs.

Waterloo Aerial Robotics Group http://ece.uwaterloo.ca/~warg In the words of the website, The Waterloo Aerial Robotics Group [at the University of Waterloo in Canada] is a team of engineers who are developing

a series of fully autonomous vehicles (both air and ground). The goal is to have a fleet of robots that can work cooperatively toward some predefined goal without the slightest bit of help from any human crew. SV

CONTACT THE AUTHOR Gordon McComb can be reached via email at [email protected]

Tough Enough for Varsity Engineering Games GEARS-IDS™™

• Computational Learning and Motor Control Lab • Laboratory for Molecular Robotics • Robotic Embedded Systems Lab Designed & Made in USA

• Polymorphic Robotics Laboratory (reconfigurable robotics)

University of Toronto Robotics & Automation www.mie.utoronto.ca/labs/ral University of Toronto Robotics and Automation Laboratory and Mechatronics Laboratory.

USC Robota Dolls www-humanoids.usc.edu/SH_ summary.html Playing with dolls at the University of Southern California. According to the website, the ROBOTA dolls are a family of mini humanoid robots. They are educational toys and can engage in complex interaction with humans, involving speech, vision, and body imitation.

USU ECE Center for SelfOrganizing and Intelligent Systems www.csois.usu.edu/index.php Robot work at the Utah State

Vexed by the Limitations of your Robot Kit? Go with the original GEARS Invention & Design System, designed by an educator for teachers and students. GEARS includes free educational resources that relate to classroom-tested, industry-grade engineering components: 8 !LL METALGEARHEADMOTORS Five times the speed, ten times the power and hundreds of times the expected life cycle of plastic servomotors. 8 0NEUMATICS Powerful and safe short-stroke cylinder, precision controlled pressure relieving regulator with gauge, electronic solenoid, shut off valve and reservoir. 8 Round Stainless Steel Axles, bronze bearings, #25 pitch sprockets and chain, Delrin gears, precision-machined aluminum wheels and stainless steel hub adapters. 8  6OLT0OWER Rechargeable battery and commercial grade charger. 8 (ARDWEARING .090 aluminum and .062 stainless structural components for building robust mechanisms. Easy to assemble #10 fasteners.

The GEARS-IDS open platform gives you the power to choose! Select additional mechanical, electrical and pneumatic components from any source. Flexibility to use various controllers – even from your existing robot kit. Utilize salvaged or surplus motors using GEARS supplied motor mounts. Add your own sensors, cameras, electronics and GPS systems to our sturdy platform.

Approved by high schools and colleges across the country. Gears Educational Systems, LLC 105 Webster Street, Hanover, MA 02339 sWWWGEARSEDSCOM

SERVO 11.2007

61

TwinTweaks.qxd

10/9/2007

11:20 AM

Page 62

THIS MONTH: Welcome to the Family

T

he holiday season is almost here — a time for family and fun. One family gathering that would be very interesting to see would be the reunion of the WowWee robotics family; the Robosapien clan. WowWee robotics has produced some of the coolest robotics toys around, and we are proud to introduce two new members of the tribe: Roboquad and Robopanda.

Surf and Turf The Roboquad looks like no identifiable creature (a Gridbug from hte movie Tron? - Ed.) It’s like a

ROBOQUAD.

62

SERVO 11.2007

four-legged, crab-like spider-like, wideeyed envoy from another world or the most mysterious corners of ours. The Roboquad is billed as the most aware of the WowWee clan thanks to an advanced vision system that relies on “deep IR scanning.” After seeing the Roboquad in action, we are inclined to agree. The bot plods around like a curious crab, and instead of relying on touch sensors like its ancestor the Robosapien, the Roboquad relies completely on its keen sense of sight. Another thing that the Roboquad has on all of its WowWee relatives is that it is by far the best dancer. With its hip tunes and flashing lights, the

ROBOQUAD ALL

BOXED UP.

Roboquad is ready for the clubs. The Roboquad’s epic personality is comprised of three elements that can be mixed and matched with three levels of intensity per element. Those three components are: awareness, activity, and aggression. Awareness refers to Roboquad’s tendency to pause, take a moment for reflection, and scan with its IR sensors. Roboquad’s impressive sensors can discern obstacles up to six feet away, and higher awareness levels will scan more often and at the maximum range. Conversely, lower awareness means rarer scans at shorter range. No matter what the Awareness setting, the Roboquad can gather an extraordinary amount of information through its sensors. It can tell if it is light or dark, and will turn its headlights on in the dark. The Roboquad can even sense movements and the shape of objects like doorways or steps. Activity refers to Roboquad’s speed and magnitude of reactions. Higher activity levels mean faster wandering and more involved responses. For

TwinTweaks.qxd

10/9/2007

11:20 AM

Page 63

Welcome to the Family example, a Roboquad set with a high activity level might react to a sensed obstacle by backing up, turning around, and completely changing course while a mellower Roboquad with a lower activity setting might plod around slowly and react to obstacles by just taking a step back. Aggression refers to the nature and intensity of Roboquad’s reactions when faced with input and obstacles. A Roboquad with a high aggression level might react by roaring and lunging forward, while a more docile Roboquad with a low aggression setting might make a discouraged sound and meekly back away. Each of these personality components have three levels of intensity that correspond to different colors on the remote — green for the first level, orange for the second, and red for the third. The personality components mainly come into play during the autonomous wandering mode, and the three elements and their levels can be recombined and reconfigured to create literally hundreds of variations in the Roboquad’s autonomous personality. The Roboquad is programmable in a way similar to other members of the WowWee family — a sequence of moves is inputted via the controller while in program mode, and at the push of a button the robot will repeat that sequence of moves. The Robosapien could run some fairly long programs even with just a maximum sequence of 15 moves. The Roboquad can remember a sequence of up to 40 moves, and it has the capability for greater customization than Robosapien in the sense that each leg can be controlled individually (or as a pair, if you like). During the programming process, the Roboquad will also demonstrate each move as inputted, which would be very helpful when creating new gaits.

Cracking the Shell of the Mysterious Spider-Crab Roboquad’s earliest ancestor, the Robosapien, was renowned far and

LOOK

INTO MY EYES

...

wide for its hackability. Intrepid tinkerers would equip their Robosapiens with everything from cameras and surround sound systems to tank tread bases. After much deliberation, we have come to the regrettable conclusion that the Roboquad is not meant to have the same hackability as its ancestors. Our first hint was the sheer number of screws that holds the Roboquad’s exoskeleton together. The Robosapien had one main body panel that was held in place by a very efficient number of screws, while the main shell of the Roboquad was held in place by over a dozen screws. Even after passing the trial of the screws, the body panels are assembled in such a way to further discourage disassembly. The main body panel is held down by the base of the neck, and the base of the neck can’t be removed unless the rest of the neck is removed. And so on. We decided to try another plan of attack, starting at the head instead of the robot’s middle. The skull of the Roboquad is held together by eight screws, all of which were extraordinarily tight. Even after removing all but a few stubborn screws, the panels of the Roboquad’s head still gave no signs at all of wanting to split open. But if we were able to open up the Roboquad, what would we find? If we were to submit to conjecture, we might make the educated guess that the Roboquad’s board is either housed just above the black box on its underbelly that holds the batteries or in the head of the beast. If we were

STRIKE A

POSE!

correct, that would mean the board is fairly small, and with the sophisticated sensor array and number of motors, we would be inclined to believe that board would be very busy indeed. This was also the case with the Roboquad’s close relative, the Roboreptile, and the busy board is not exactly hacker friendly. In the face of compelling evidence, we were forced to make the unfortunate conclusion that the Roboquad simply isn’t meant for hacking. Of course, we think that tenacious tinkerers will be able to surmount all of these obstacles and do some crazy things to the mysterious spider-crab, but the design of the bot certainly won’t make it too easy of an undertaking.

ROBOPANDA.

SERVO 11.2007

63

TwinTweaks.qxd

10/9/2007

1:31 PM

Page 64

Twin Tweaks ...

ROBOPANDA

CARTRIDGE.

Eats Shoots and Leaves The Robopanda is the other new member of the WowWee clan and it will immediately captivate audiences of all ages with its charming androgynous voice and animated movements. Robopanda is a completely interactive robotic companion — so interactive that any activities cannot be one sided because there is no remote to accompany the robotic bear. The Robopanda interacts with its environment through touch sensors on its feet, paws, back, belly, and head, and a sound sensor. It takes a bit of practice to become an expert as to where all of Robopanda’s touch sensors are, but in just a short while the moments of head scratching that follow the command “rub my belly if you want to hear a story” will be a thing of the past. As a nice safety feature, the Robopanda sports several pinch sensors around its arms and neck to prevent hapless children from pinching their fingers. And even if kids don’t grasp the danger of the pinch points, they should certainly learn their lesson from Robopanda’s tortured cries of “please don’t hold my neck that way!”

Panda-monium All WowWee toys are billed as a fusion of technology and personality, and the Robopanda is downright emotive. Expressive blue eyes, moving eyebrows, and an animated voice with real human-like inflections are just the

64

SERVO 11.2007

THE ROBOPANDA

SWITCH!

OOOH! AAAH!

start. The Robopanda displays a wide array of emotions and moods that range from excitedly telling stories and jokes to fitfully demanding his Teddy. And while the Robopanda’s expressive voice alone would have been plenty to define all of these moods, it completes the picture with playful movements of its paws and ears. The Robopanda can operate in any of three different modes. Menu Mode allows you to choose what Robopanda does, whether it is to tell you stories or jokes or play games like “Panda Says.” Robopanda’s stories are packed full of information about other animals, and in addition to being educational, the Robopanda tells them with enough zeal to be undeniably entertaining.

Teaching an Old Panda New Tricks Training Mode allows you to teach Robopanda tricks. Teaching the boisterous bear new tricks can be achieved in just a few easy steps. First, you have to choose where Robopanda will remember the trick by pressing one of his touch sensors. Later on when you want Robopanda to perform a trick (or if it prompts you), all you have to do is press the sensor to which the trick is assigned. Second, you have to teach Robopanda the trick itself. What a trick actually is to Robopanda is basically a remembered movement. You could make Robopanda flap its arms like a bird or give a karate chop — all you have to do is physically move Robopanda’s arms accordingly,

and it will remember the trick. When we first started training our Robopanda, we were a bit wary about how much force we could use in manipulating the Panda’s paws for the trick. The bear is surprisingly robust, and being too gentle will not allow Robopanda to register the trick. Nailing down the complexity of the movements that the Robopanda can remember is also a trial and error process, but overall, the tricks feature is a fun feature that paints that much more of a complete picture of interactivity. And don’t fret if you can’t remember these steps — Robopanda will cheerily guide you through them every time.

Panda Expressive Robopanda’s premier mode is “Friend Mode,” and we have to admit that the irrepressible robot makes an entertaining pal indeed. You could literally spend hours listening to Robopanda excitedly recount its dreams, demand foot massages, and spout seemingly endless bits of panda related trivia. We didn’t know that Pandas are nicknamed Great Bear Cats because of their cat-like eyes, and we had never heard of Red Pandas before either. The Robopanda’s animated conversations will at the very least give you an edge in Trivial Pursuit: Jungle Edition, and at the most pique your curiosity about the Great Bear Cats to the extent that you continue researching the fascinating animals long after Robopanda falls asleep (which it will do from time to time). We looked up the Red Panda on Wikipedia, and we can assure you that it certainly fits the description given by Robopanda (“a lot smaller and a lot redder than I am”), even though Robopanda forgot to mention how overwhelmingly precious it is. Perfectly timed reactions give the Robopanda the convincing illusion of sophisticated artificial intelligence. It will ask you questions, pause for just long enough for you to give an answer, then react in what always seems to be a delightfully appropriate way.

TwinTweaks.qxd

10/9/2007

11:21 AM

Page 65

Welcome to the Family Friend Mode is also where Robopanda will demand its Teddy either for hugs or a game of peek-aboo. The Robopanda can actually identify its Teddy through radiofrequency identification, and the chip was so seamlessly implanted in the stuffed animal that we would have to cut open the hapless bear to find it — though that does leave the compelling possibility of switching Teddy’s chip with a banana or a roll of toilet paper and having Robopanda play peek-a-boo with that instead. Robopanda will rarely repeat itself while going through its vast repertoire, but that repertoire is essentially limited. Such a limitation, however, does not pose much of a threat to Robopanda’s fun factor because its repertoire is actually cartridge based, and the bear bot comes with two cartridges. Each cartridge comes with its own set of stories, jokes, games, and panda trivia. While this is exciting simply for the promise of hours of panda fun, it also opens the door to the possibility of new cartridges being released in the future. These future cartridges could contain even more stories and jokes and panda trivia, and ensure that the fun of the Robopanda never becomes obsolete. The other side of the cartridge potential is the possibility of hacking the panda. Perhaps custom external devices could be attached via the connector for the cartridges. But what about more traditional hacking by opening up the Robopanda itself? The Robopanda, unfortunately, also seems to have been designed to discourage hacking. On every segment of the panda’s shell, we were able to remove all but one or two screws, with the troublesome screws simply being to far inset into the shell to reach with our screwdriver. And even if a hacker was able to crack open the panda (as terrible as that sounds), we think they would have to be very careful with their hacks indeed. Robopanda’s primary functionality is in its speech, so any electronics additions should be careful to avoid compromising that. Perhaps a mechanical hack would be best, like adding wheels or treads to

the Robopanda’s feet so it could really move around — ever so often Robopanda will declare its intention to crawl around, but unless it is on the perfect surface it amounts to little more than crawling in place.

From Homosapien to Robosapien and Beyond One thing that is true of the Robopanda, Roboquad, and all of the WowWee family is that they are very sophisticated in their simplicity. The Roboquad is particularly surprising. Its crab-like walk and multitude of variants surely seem to be the result of a complicated arrangement of motors and gears, but in reality the endearing gaits are achieved simply by one motor per leg (the whole bot has six motors total; four in the legs and two in its neck). The exceptional cleverness that makes the Roboquad so full of personality is in the design of the bot itself. It’s not surprising that this bot was designed by Mark Tilden, the designer of the Robosapien and father of BEAM (Biology, Electronics, Aesthetics, and Mechanics) Robotics. In their achievement of such a high caliber of sophistication and coolness, the Roboquad and all of the WowWee robots seem to be guided by similar principles as those guiding BEAM robotics — simplicity, efficiency, and elegance. Many BEAM robots use solar power, and even though the Roboquad and Robopanda run on batteries the efficiency of their design is in the same

ROBOPANDA TEDDY.

spirit of economy. They use a minimal number of motors to achieve complicated movements, and the bots are designed to take advantage of reflexive motion to use less energy. BEAM robots are also characterized by simple electronics that use logic circuits as opposed to microprocessors. We weren’t able to get a look at the electronics of the Roboquad or Robopanda, but if their relatives are any indication, we imagine that they are paragons of simplicity and efficiency. The A in BEAM is aesthetics, and the WowWee bots have always done well in this department. From the sleek color schemes to the expressive faces, the Roboquad and its ilk are works of art as much as robotic toys. The final element is mechanics, and with sparse usage of motors and creative reflexive motion, the WowWee bots are mechanical works of art, as well. Complex movements can be achieved in a complicated way, but achieving it in a simple way is far greater of a challenge. The Roboquad achieves its spider-crab gait though four motors instead of the eight or 10 it easily could have used. And the real trick with this whole thing is that the Roboquad doesn’t sacrifice any of its realism or charm by using fewer motors — that is good mechanical design. As for the Robopanda, while it might not meet all of those conditions of BEAM robotics, we think that it will still make an astounding contribution to the world of robotics as a whole for crossing that difficult boundary

THE

FAB TRIO.

SERVO 11.2007

65

TwinTweaks.qxd

10/9/2007

11:21 AM

Page 66

Twin Tweaks ... Recommended Websites For more information, go to: www.roboquadonline.com www.robopandaonline.com www.wowwwee.com between robotic toy and robotic companion. Many futurists, roboticists, and regular folks imagine that in the near future robots will be a part of our everyday lives. For many others, however, it is hard to visualize robots becoming so integrated into society. A Roomba scurrying across the floor seems another thing entirely from an Asimo-like humanoid helping you carry the groceries. We think the disconnect comes from the expectation that robots performing human-like tasks should be able to interact with us in human-like ways. We also think that the Robopanda bridges that divide. Then, what is the task that the

Robopanda fulfills so effectively that demands such human-like finesse? The deceptively difficult task is companionship. The Roboquad, with all of its funky sounds and flashing lights, is a toy and not a companion. It can’t interact with us like a person can, through conversation and delightful reactions. The Robopanda does these things, and even though it is in the form of a lovable bear, the essence of its success lies in the fact that it is indeed lovable.

Descending the Family Tree As the descendants of the Robosapien become more numerous, they also seem to become more sophisticated and self contained. A look at the evolution of the WowWee remote control is indicative of this trend. The Robosapien’s controller is populated by copious amounts of buttons even though the number of buttons is mitigated by the color coded levels.

STEER WINNING ROBOTS WITHOUT SERVOS!

Extreme Robot Speed Control! 6 6 6 6 6 6 6

$399

14V - 50V - Dual 80A H-bridges - 150 A+ Peak! Adjustable current limiting Temperature limiting Three R/C inputs - serial option Many mixing options - Flipped Bot Input Rugged extruded Aluminum case 4.25" x 3.23" x 1.1”

$119.99

$79.99 $29.99 Scorpion Mini

6 2.5A (6A pk) H-bridge 6 5V - 18V 6 1.25“ x 0.5” x 0.25”

6 6 6 6

Scorpion HX

Scorpion XL

Dual 2.5A (6A pk) H-bridges Plus 12A fwd-only channel 5V - 18V 1.6“ x 1.6” x 0.5”

6 Dual 13A H-bridge 45A Peak! 6 5V - 24V 6 2.7“ x 1.6” x 0.5”

Introducing Dalf 6 6 6 6 6 6 6

P

Closed-loop control of two motors Full PID position/velocity loop Trapezoidal path generator Giant Servo Mode! $250 PIC18F6722 CPU C source for routines provided See www.embeddedelectronics.net

H-bridges: Use with Dalf or with your Micro/Stamp Simple-H

OSMC 6 6 6 6

Monster power! 14-50V 160A! 3.15”x4.5”x1.5” 3 wire interface

$79

6 6 6 6

6-28V 25A! 2.25”x2.5 ”x0.5” 3 wire interface current & temp protection

MADE IN

www.robotpower.com

THE USA

Phone: 253-843-2504
[email protected]

66

SERVO 11.2007

The Roboreptile and the Roboquad’s remotes are much more sparsely populated, and the Robopanda has no remote at all. The hackability of the bots also follows this trend. The Robosapien was a hacker’s dream, but now the Roboquad and Robopanda seem to specifically discourage hacking. This trend might seem like a negative thing to many tinkerers, but the other side of this coin is an exciting advancement in the robotic field, both technologically and socially. Technologically because these are advanced robots with sophisticated sensor arrays that are widely available, and socially because they make the seemingly science fiction inspired vision of a future filled with robotic companions seem much more realistic. These are toys cool enough to tear kids away from the video games, and they are robots sophisticated enough to garner the admiration of hobbyists and professionals alike. SV

erform proportional speed, direction, and steering with only two Radio/Control channels for vehicles using two separate brush-type electric motors mounted right and left with our mixing RDFR dual speed control. Used in many successful competitive robots. Single joystick operation: up goes straight ahead, down is reverse. Pure right or left twirls vehicle as motors turn opposite directions. In between stick positions completely proportional. Plugs in like a servo to your Futaba, JR, Hitec, or similar radio. Compatible with gyro steering stabilization. Various volt and amp sizes available. The RDFR47E 55V 75A per motor unit pictured above. www.vantec.com

Order at (888) 929-5055

SV-GiftSubAd2007.qxd

10/10/2007

7:13 PM

Page 67

Give the gift that lasts all year!

Go to

www.servomagazine.com to enter your gift order!

Be sure to use promotion code Y7WXMS when ordering.

U.S. Prices 1 Yr – $24.95 2 Yr – $45.95 3 Yr – $65.95

Subscribe online at: www.servomagazine.com or call: 877-525-2539 (toll free) / 818-487-4545 (outside US)

DifferentBits.qxd

10/9/2007

11:17 AM

Page 68

DIFFERENT BITS

NEURAL NETWORKS FOR THE PIC MICROCONTROLLER PART 3 — HEBBIAN LEARNING by Heather Dewey-Hagborg

In September’s column, we looked at feedforward neural networks and backpropagation and discussed their uses for solving linearly inseparable problems. This month, we are going to look at a more biologically plausible class of neural network models based on Hebbian learning. This article will assume that you have been following along with the series and are familiar with terms like input sum, activation, weights, and learning function.

H

ebbian learning is based on a simple and intuitive principle: connections between neurons that tend to fire together increase in strength. The converse of this is also true: Neurons that rarely fire together have weak connections. This means that neurons become “associated” and assist in each other’s activity. Hebbian learning allows neural networks to form simple behavioral conditioning. By repeatedly presenting a set of input and output patterns together to the network, it will learn over time to associate the input and other similar inputs with the output. The most basic Hebbian learning networks have only two layers of neurons, inputs, and outputs. They use the same linear, hard-limiting activation function as the perceptron neural network from July’s column: activation = sign(input sum) In other words, if the input sum is > 0, the activation is 1; otherwise, the activation is -1. The input sum is computed in the same way as always:

68

SERVO 11.2007

input sum = (input0 * weight0) + (input1 * weight1) + … And so on, for as many inputs as the neuron has. Learning can be either supervised or unsupervised. So far in this column, we have discussed only supervised models of learning where we play the role of teacher and effectively tell the neural network the correct answer repeatedly until it gets it right. Hebbian learning can involve a teacher or can work on its own without anyone telling the network what is right and wrong. Both forms of learning go through the same cycle of sensing and updating connection strengths, but the unsupervised model computes weight updates based on what values are present together, whereas the supervised version computes based on what the output should be. The learning rule is as follows: weight change = learning constant * desired output * input value Or, in the unsupervised model:

weight change = learning constant * neuron A activation * neuron B activation As you can see, connection strength will only increase between correlated neurons; neurons with activations of the same sign. For example, if we want an output neuron to have an activation of 1 and it has two inputs — one input activated to -1 and the other input activated to +1 — the weight change will be positive between the neurons with the same sign, and negative between the neurons with opposite signs. EX: weight change = 0.2 * 1 * -1 weight change = -.2 weight change = 0.2 * 1 * 1 weight change = .2 Hebbian learning is often used in neural networks that model psychological theories. The most common example of this is a form of Pavlovian classical conditioning where an animal (or robot) learns to associate two

DifferentBits.qxd

10/9/2007

11:17 AM

Page 69

DIFFERENT BITS

FIGURE 1. Circuit schematic.

SERVO 11.2007

69

DifferentBits.qxd

10/9/2007

11:17 AM

Page 70

DIFFERENT BITS FIGURE 2. Hebbian robot neural network diagram.

and Jacques Ludik (ISBN: 0-521-57163-4).

Microcontroller Implementation

stimuli, such as the sound of a bell and receiving food. Hebbian learning is also used in more complex hybrid neural network models to simulate a variety of psychological conditions and principles. For more on this subject, check out Computational Explorations in Cognitive Neuroscience by Randall C. O’Reilly and Yuko Munakata (ISBN: 0-262-65054-1) and Neural Networks and Psychopathology by Dan J. Stein

BILL OF MATERIALS • Breadboard For microcontroller circuit: • PIC16F877A microcontroller and programmer • 20 MHz ceramic resonator with built-in capacitors (or equivalent) • One pushbutton (for reset) • One 10K resistor (for reset) • One light emitting diode (status light) • One 220 ohm resistor (for the LED) For serial communication: • RS-232 level shifter (MAX233 or similar) • 1 µF capacitor (if using MAX233) • Serial cable with receive pin available for breadboard use For input: • Four one-bit digital input sensors, i.e., pushbutton/whisker switches, proximity detectors, etc., and coordinating circuitry For output: • H-bridge chip for two motors, i.e., L293D • Two motors • Two capacitors • Separate power supply for motors

70

SERVO 11.2007

In this example, we will create a breadboard prototype for a robot that learns to map sensor inputs to outputs for controlling motors. The program is split into two phases. First, the robot learns the input/output mappings you provide. Second, it begins a senseact cycle where it reads the values of its sensors, processes those through its neural network, and passes the output values to the inputs of an H-bridge for motion. At the end of this section, we will discuss how you could expand the example to create a robot that learns from supervised instructions received in real-time via a remote control. Circuit Wire up the PIC chip and MAX233 chip as usual (see July column for more detailed instructions and pictures). For my breadboard prototype, I just used four pushbutton switches to simulate thresholded proximity sensors. I connected the switches between the associated microcontroller pin and +5V and then connected pull-down resistors between that microcontroller pin and ground. For motors, I used two Solarbotics GM10 geared pager motors running at three volts DC. For wiring details, reference my schematic in Figure 1. Feel free to use whatever combination of sensors and motors/motor drivers you prefer; the point here is the learning algorithm not the hardware. Code The code (which you can download from the SERVO Magazine website at www.servomagazine. com) is broken down into three main sections: setup/main loop, learning, and motor commands. Let’s look at

the learning functions first. • sign — Simply returns the sign (positive 1 or negative 1) of the floating point argument. • getActivation — This function computes the output neuron activations based on the current set of inputs and connection strengths. • learn — This function applies the Hebbian learning rule described above to the neural network connections based on the current inputs and desired outputs. weight change = learning constant * desired output * input value • calculateError — This function computes the root mean squared error of the network’s learning each iteration and prints it to your serial monitor. This gives you a sense of how quickly the network is learning and provides a troubleshooting measure. The motor control functions are relatively self explanatory. Each motor has an enable pin and two direction pins that specify whether it is turning forward or backward; the motor commands set these three pins appropriately. • motor1_forward, motor2_forward — Enable the motor and set one direction pin high and the other low. • motor1_backward, motor2_back ward — Enable the motor and flip the direction pin bits to the reverse direction. • motor1_stop, motor2_stop — Set the enable pin low to disable the motor. Finally, let’s look at the setup and main loop functions. • setup — Disable all the functions we are not using, print a “Hello!” message out the serial port, and turn on the status LED. This function helps us know that basic functions like pin I/O and serial transmission are working properly.

DifferentBits.qxd

10/9/2007

11:17 AM

Page 71

DIFFERENT BITS • main — First, we run the setup command. Second, we initialize the connections array with small floating point random weights. Third, the neural network goes through its learning phase which generally takes less than 50 iterations. It prints what iteration it is out the serial port and then randomly chooses an input set for training. It finds its activation, runs the learning function, and computes the current root mean squared error. When the learning phase is complete, the program switches to an infinite loop of sensing, processing the input array through the neural network, and activating the output motors. The public array outs[] holds the motor command outputs from the neural network. A value of +1 means motor forward, and a value of -1 means motor backward. After programming, you should see the following behavior from your circuit: 1) The status LED turns on and the “Hello!” message prints to your serial monitor. 2) Initial weight values print out the serial port. 3) Documentation of the learning process prints out showing you what iteration and error level the network is at. This completes almost too quickly for you to notice. 4) The motors will start turning forward and debugging information will print out the serial port telling you what input values are read and what output values are calculated. Evaluate the neural network. Test it by changing the sensor values and seeing how it responds. Does it act the way you taught it to? How does it respond to ambiguous input situations that you did not explicitly program in? Does it generalize well?

Next Steps ... Having a robot that learns and is

able to generalize from what it has learned to gracefully handle new situations is great. But it would be nice if you didn’t have to imagine a bunch of situations beforehand and program them explicitly. One solution I have imagined for this is using a remote control as an on-the-fly teacher for an embodied robot. The robot would start with the same randomly initialized neural network as in this program, but rather than defining a set of inputs and correct responses, the robot would go through a learning cycle every time it received a remote control instruction; the instructions would function as behavioral corrections. This is an appealing idea because it allows the robot to learn in a more natural way, similar to a pet. To implement this process, you just need to change the main function by deleting the learning phase and adjusting the loop to check for serial data after each move. If a command has been received, the robot adjusts its position and runs the learn function with the new information loaded. The learn function needs to be revised to calculate the delta value based on the actual output neuron values rather than desired values: delta = constant*(float) outs[k]* (float) ins[j]; And finally, I would include a serial interrupt function so that you don’t have to lose instructions waiting for a byte that never arrives: #int_rda void serial_isr() { cmd=getc(); putc(cmd); new=1; }

Have fun experimenting with these concepts! I would love to hear from any of you that attempt to implement or modify these algorithms in your own projects. If you want to share your successes (or frustrations) with me, drop me a line at heather@servo magazine.com. SV SERVO 11.2007

71

WebstoreNov07.qxd

10/10/2007

9:51 AM

Page 72

The SERVO Webstore Attention Subscribers ask about your discount on prices marked with an * Insectronics

by Karl Williams

This complete project book delivers all the stepby-step plans you need to construct your own six-legged insect-like robot that walks and actually responds to its environment. Using inexpensive off-the-shelf parts hobbyists can “build a better bug” and at the same time have loads of fun honing their knowledge of mechanical construction, programming, microcontroller use, and artificial intelligence. $19.95

Mechanisms and Mechanical Devices Sourcebook

2007 CD Rom

Pre-order before December 10th and get free shipping (U.S. Only) $24.95

Order other CDs with your 2007 copy of SERVO or NV (of any combination) and pay only $21.95 each. Plus get the free shipping! All orders will ship mid December.

123 Robotics Experiments for the Evil Genius by Myke Predko

If you enjoy tinkering in your workshop and have a fascination for robotics, you’ll have hours of fun working through the 123 experiments found in this innovative project book. More than just an enjoyable way to spend time, these exciting experiments also provide a solid grounding in robotics, electronics, and programming. Each experiment builds on the skills acquired in those before it so you develop a hands-on, nuts-and-bolts understanding of robotics — from the ground up. $25.00

Robot Builder’s Sourcebook

CNC Robotics

by Gordon McComb

Fascinated by the world of robotics but don’t know how to tap into the incredible amount of information available on the subject? Clueless as to locating specific information on robotics? Want the names, addresses, phone numbers, and websites of companies that can supply the exact part, plan, kit, building material, programming language, operating system, computer system, or publication you’ve been searching for? Turn to Robot Builder’s Sourcebook — a unique clearinghouse of information that will open 2,500+ new doors and spark almost as many new ideas. $24.95

72

SERVO 11.2007

by Neil Sclater / Nicholas Chironis

The fourth edition of this inventioninspiring engineering resource covers the past, present, and future of mechanisms and mechanical devices. You’ll find drawings and descriptions of more than 2,000 components that have proven themselves over time and can be incorporated into the very latest mechanical, electromechanical, and mechatronic products and systems. Overviews of robotics, rapid prototyping, MEMS, and nanotechnology, along with tutorial chapters on the basics of mechanisms and motion control, will bring you up-to-speed quickly on these cutting-edge topics. $89.95

FIRST Robots: Aim High

by Vince Wilczynski / Stephanie Slezycki

This book looks at 30 different robot designs all based on the same chassis, and provides indepth information on the inspiration and the technology that went into build! NEW ing each of them. Each robot is featured in 6-8 pages providing readers with a solid understanding of how the robot was conceived and built. There are sketches, interim drawings, and process shots for each robot. $39.95

by Geoff Williams

CNC Robotics gives you step-by-step, illustrated directions for designing, constructing, and testing a fully functional CNC robot that saves you 80 percent of the price of an off-the-shelf bot — and that can be customized to suit your purposes exactly, because you designed it. Written by an accomplished workshop bot designer/builder, this book gives you everything you need. $34.95 We accept VISA, MC, AMEX, and DISCOVER Prices do not include shipping and may be subject to change.

WebstoreNov07.qxd

10/10/2007

9:52 AM

Page 73

To order call 1-800-783-4624 Building Robots with LEGO Mindstorms NXT by Mario Ferrari, Guilio Ferrari

! NEW 7 0 N JU The Ultimate Tool for MINDSTORMS® Maniacs, the new MINDSTORMS kit has been updated to include a programming brick, USB cable, RJ11-like cables, motors, and sensors. This book updates the robotics information to be compatible with the new set and to show how sound, sight, touch, and distance issues are now dealt with. $39.95

Forbidden LEGO

NEED MORE INFO? Go where SERVE’ goes! The online store @ www.ser v omagazine .com

by Ulrik Pilegaard / Mike Dooley Build the Models Your Parents Warned You ! Against. NEW Forbidden LEGO introduces you to the type of freestyle building that LEGOs master builders do for fun in the back room. Using LEGO bricks in combination with common household materials (from rubber bands and glue to plastic spoons and ping-pong balls) along with some very unorthodox building techniques, you’ll learn to create working models that LEGO would never endorse. $24.95

Robot Builder’s Bonanza Third Edition

by Gordon McComb / Myke Predko

Everybody’s favorite amateur robotics book is bolder and better than ever — and now features the field’s “grand master” Myke Predko as the new author! Author duo McComb and Predko bring their expertise to this fully-illustrated robotics “bible” to enhance the already incomparable content on how to build — and have a universe of fun — with robots. $27.95

Take This Stuff and Hack It!

Great ! a Gift Ide ! 5 $19.9

by Dave Prochnow

Transform common household items into really cool stuff. You don't need to be an electronics genius to get started turning everyday items into high-performing wonders. With how-to guru Dave Prochnow's step-by-step directions and fully illustrated plans, even beginners can hack their way to a high-tech home, cooler toys, and less yard work. Certain to fire your imagination and start you plotting new, original, and even more creative wonders you can make from ordinary household items, Take This Stuff and Hack It! is the perfect gift for your inner inventor. $27.95

SERVO Magazine T-Shirts

From HomoSapien to RoboSapien

For men and women

Before R2D2 there was R1D1

Get your very own limited-edition SERVO Magazine T-shirt. Shirts come in sizes S, M, L, and are available in either black or white. All shirts are 100% preshrunk cotton. SERVO 11.2007

73

WebstoreNov07.qxd

10/10/2007

9:53 AM

Page 74

To order call 1-800-783-4624 Or order online www.servomagazine.com NEW! THE SERVO STORE NOW OFFERS DVDS!

The Day the Earth Stood Still An alien (Klaatu) with his mighty robot (Gort) land their spacecraft on Cold War-era Earth just after the end of World War II. They bring an important message to the planet that Klaatu wishes to tell to representatives of all nations. Directed by: Robert Wise Actors: Michael Rennie, Patricia Neal Run Time: 92 minutes $14.95

Blade Runner (The Director’s Cut) (1982)

In a cyberpunk vision of the future, man has developed the technology to create replicants, human clones used to serve in the colonies outside Earth but with fixed lifespans. In Los Angeles, 2019, Deckard is a Blade Runner, a cop who specializes in terminating replicants. Directed by: Ridley Scott Rated: R — Violence $19.95

Forbidden Planet: 50th Anniversary Edition

A starship crew goes to investigate the silence of a planet’s colony only to find two survivors and a deadly secret that one of them has. Dr. Morbius and his daughter Altaira have somehow survived a hideous monster which roams the planet. Directed by: Fred M. Wilcox Run Time: 98 minutes $26.95

BACK ROOM SPECIALS

The Official Robosapien Hacker's Guide by Dave Prochnow

The Robosapien robot was one of the most popular hobbyist gifts of the 2004 holiday season, selling approximately 1.5 million units at major retail outlets. The brief manual accompanying the robot covered only basic movements and maneuvers — the robot's real power and potential remain undiscovered by most owners — until now! This timely book covers all the possible design additions, programming possibilities, and "hacks" not found any place else. $24.95 Sale Price $14.95 Pub date: August 17, 2005

Concise Encyclopedia of Robotics by Stan Gibilisco

Written by anaward-winning electronics author, the Concise Encyclopedia of Robotics delivers 400 up-to-date, easy-to-read definitions that make even complex concepts understandable. Over 150 illustrations make the information accessible at a glance and extensive cross-referencing and a comprehensive bibliography facilitate further research. $19.95 Sale Price $9.95 “Only 2 Left” Pub date: November 12, 2002

Robotics Demystified

by Edwin Wise YOU DON'T NEED ARTIFICIAL INTELLIGENCE TO LEARN ROBOTICS! Now anyone with an interest in robotics can gain a deeper understanding — without formal training, unlimited time, or a genius IQ. In Robotics Demystified, expert robot builder and author Edwin Wise provides an effective and totally painless way to learn about the technologies used to build robots! $19.95 Sale Price $13.95 Pub Date October 20, 2004

H-racer-and-Hydrogen-station

SERVO'S CUSTOM EMBROIDERED HAT

With this kit, you now can see and feel the future of energy generation in your own hands! This is one good looking hat, with its stone washed blue cap and bright yellow/gold stitching. There's no better way to show everyone that you're well endowed with a bigger - than - average brain then by proudly displaying your affiliation with SERVO Magazine. $ 14.95*

74

SERVO 11.2007

Recently named as one of the Best Inventions of 2006 by Time Magazine, the H-racer is now the best selling fuel cell product in the world. The H-racer is a micro-version of what engineers and scientists have been dreaming about for real cars: combining hydrogen with oxygen to generate a DC current to power an electric motor. Unlike a gas-powered car engine, the only byproducts of this electrochemical process are electricity, heat, and pure water. With the H-racer, you can witness the power of new energy technology in the palm of your hand. Horizon has created this unique, patented miniature fuel-cell car and hydrogen refueling station. All you need to do is add water. Only $115.00 plus S/H

Mind-Feed Nov07.qxd

10/10/2007

4:35 PM

Page 75

continued from page 7

Response: You are quite right in pointing out that traditional differentials use a twin axle, and that this does result in a more efficient transmission of power to the back wheels while the vehicle is turning. This is actually the way the new Vex differential kit is designed to be used, as it comes with all of the requisite components to build a diff just as you mentioned. Since we like to do things a bit differently, however, we chose to show how the parts could be used in another way and, as we stated in the article, the fixed diff is an old school strategy to maximize power transmission to the wheels while the vehicle is moving forward.The fixed diff does have the potential to pose serious problems for turning, but we think the slightly off-kilter geometry of the front end was the primary culprit — as we drove the vehicle, we could see that it was the front wheels that were scrubbing and not the back wheels that were hopping.Thanks for reading! — Bryce and Evan Woolley Dear SERVO: Alex Dirks of CrustCrawler wrote to you concerning what he felt were incorrect statements that I made in my July Then & Now column concerning servos used in many experimenter’s robots — the types used with radio-controlled model airplanes to control various wing and rudder surfaces. He referred to the following statements that I made concerning their use with robot arms:

mechanisms built into ANY standard servo today with the exception of the AX-12+ and a few specialized servos used in biped type robots.” I have since spoken with Alex on the subject and we see eye-to-eye. It turns out that we are both right. Here is a definition that I saw on the PC Magazine website: (a servo is) “An electromechanical device that uses feedback to provide precise starts and stops for such functions as the motors on a tape drive or the moving of an access arm on a disk.” By nature, a servo is a feedback device and it uses the internal pot to feedback positional data to the internal circuitry of the servo. Because of this feedback, the model aircraft servo will continually try to move the servo’s output shaft to the position that the external microcontroller requested, and will only stop ‘trying’ when that point is reached. A stepper motor, by contrast, will receive a series of pulses to move a certain number of degrees, but may be jammed, and could stop at a point far from where the microcontroller requested it to be.This positional feedback is what I feel is an advantage over a stepper motor. Alex was correct in stating that a typical model airplane servo has no feedback to the outside world, including the controlling microcontroller. The new Robotis AX-12+ and a few high-end servos do, however, but not the typical hacked servo we read about. — Tom Carroll

1) “The advantage of using R/C servos is the positional feedback.” 2) “Potentiometric feedback, as in R/C servos, allows the controlling computer to know where each joint is positioned.” Alex countered with the following: “There are no feedback THE OWNERSHIP, MANAGEMENT, AND CIRCULATION STATEMENT OF SERVO MAGAZINE, Publication Number: 1546-0592 is published monthly. Subscription price is $24.95. 7. The complete mailing address of known office of Publication is T&L Publications, Inc., 430 Princeland Ct., Corona,Riverside County, CA 92879-1300. Contact Person: Tracy Kerley. Telephone: (951) 371-8497. 8. Complete Mailing address of Headquarters or General Business Office of Publisher is T&L Publications, Inc., 430 Princeland Ct, Corona, CA 92879. 9. The names and addresses of the Publisher, and Associate Publisher are: Publisher, Larry Lemieux, 430 Princeland Ct., Corona, CA. 92879; Associate Publisher, Robin Lemieux, 430 Princeland Ct., Corona, CA 92879. 10. The names and addresses of stockholders holding one percent or more of the total amount of stock are: John Lemieux, 430 Princeland Ct., Corona, CA 92879; Larry Lemieux, 430 Princeland Ct., Corona, CA 92879; Audrey Lemieux, 430 Princeland Ct., Corona, CA 92879. 11. Known Bondholders, Morgagees, and other security holders: None. 12. Tax Status: Has not changed during preceding 12 months. 13. Publication Title: SERVO Magazine 14. Issue Date for Circulation Data: October 2006-September 2007. 15. The average number of copies of each issue during the proceeding twelve months is: A) Total number of copies printed (net press run); 11,581 B) Paid/Requested Circulation (1) Mailed Outside County subscriptions: 6,069 (2) Mailed In-County subscriptions: 0 (3) Paid Distribution Outside the Mail including Sales through dealers and carriers, street vendor, and counter sales and other paid distribution outside USPS: 1,644 (4) Paid Distribution by other classes of mail through the USPS: 0; C) Total Paid Distribution: 7,713; D) Free or Nominal Rate Distribution by mail and outside the mail (1) Free or Nominal Rate Outside-County Copies: 64 (2) Free or Nominal Rate In-County Copies: 0 (3) Free or Nominal Rate Copies Mailed at other classes through the USPS: 0 (4) Free or Nominal Rate Distribution Outside the mail: 898; E) Total Free or Nominal Rate Distribution: 962; F) Total Distribution: 8,675; G) Copies not distributed: 2,906; H) Total: 11,581; Percent paid circulation: 88.91%. Actual number of copies of the single issue published nearest the filing date is September 2007; A) Total number of copies printed (net press run) 10,544; B) Paid/Requested Circulation (1) Mailed Outside County subscriptions: 5,976 (2) Mailed In-County subscriptions: 0 (3) Paid Distribution Outside the Mail including Sales through dealers and carriers, street vendor, and counter sales and other paid distribution outside USPS: 1,575 (4) Paid Distribution by other classes of mail through the USPS: 0; C) Total Paid Distribution: 7,551; D) Free or Nominal Rate Distribution by mail and outside the mail (1) Free or Nominal Rate Outside-County Copies: 0 (2) Free or Nominal Rate InCounty Copies: 0 (3) Free or Nominal Rate Copies Mailed at other classes through the USPS: 0 (4) Free or Nominal Rate Distribution Outside the mail: 300; E) Total Free or Nominal Rate Distribution: 300; F) Total Distribution: 7,851; G) Copies not distributed: 2,693; H) Total: 10,544; Percent paid circulation: 96.18%. I certify that these statements are correct and complete. Larry Lemieux, Publisher - 9/28/07.

SERVO 11.2007

75

Appetizer.qxd

10/9/2007

2:16 PM

Page 76

The Black Widow Contest Winner: ITESO’s Myoelectric Prosthetic Hand by Bryan Bergeron

T

he winning entry for Freescale Semiconductor’s recent Black Widow $10,000 Design Challenge — a myoelectric prosthetic hand — is the product of a team of electrical engineering students from ITESO graduate school in Guadalajara, Mexico. This article, based on an interview with team members, introduces the concept of yoelectric control and offers insight into the challenges inherent in an expenseand time-limited robotics project.

The Black Widow Contest The judges of the Black Widow PHOTO 1. Prosthetic hand connected via surface electrodes to the operator’s arm.

Challenge considered entries on the basis of creativity, design efficiency, technical complexity, number of Freescale devices used, and overall application innovation and usefulness. The contest required entrants to incorporate the MC9S08QG, MC9S08QD, or MC9RS08KA families of microcontrollers in their design. While the ITESO team received $10,000 and considerable publicity, all entrants received free technical training and a discounted price for development tools supporting Freescale’s S08 eight-bit microcontrollers. Registration or the design challenge closed in April 2007, with more than 775 submissions. For details of contest rules and information on the next Black Widow Contest, see www.freescale.com/blackwidow.

Winning Design The winning design — shown in PHOTO 2. Close-up of hand construction.

Photos 1 and 2 — is a myoelectric prosthetic hand. The hand can open, close, and rotate clockwise and counterclockwise, as dictated by the user. The unique aspect of the design is the myoelectric control, which is based on electrical activity in the muscles in the forearm used for gross hand movement. Although the design could have relied on surface electrodes placed over the smaller muscles within the hand, using four pairs of electrodes over muscles of the forearm enables the robotic hand to be used by amputees. The front-end circuitry detects, amplifies, and conditions the myoelectric signals before analog-todigital conversion. The digital signals are then interpreted by the microcontroller, which controls the two motors in the hand. A pair of Freescale MC33887 H-bridges are used to drive the hand motors. Grip strength is controlled manually by a potentiometer, enabling the hand to apply appropriate force to the object being grasped. Total development cost for the hand was approximately $1,800.

Lessons Learned According to the ITESO team leader, Alan Collins, the greatest challenges associated

76

SERVO 11.2007

Appetizer.qxd

10/9/2007

2:16 PM

Page 77

with developing the prosthetic hand were limited time and team management. Time was limited because the students in the team — all of whom were enrolled in a class in Biomedical Electronics — were in their last semester of studies. Approximately three months were spent in design and development, with only one month for implementation and design changes. While working under the time constraints of a school semester were trying, the greatest hurdles were associated with project and team management. Team members had to put their individual egos aside, agree to a shared vision, and work as a true team instead of seven independent designers. In the end, this may have been more important than the specifics of the technology.

Winning Team The winning team, a group of seven electrical engineering students from ITESO graduate school at the Universidad Jesuita de Guadalajara

PHOTO 3. ITESO team members, left to right: Missael Maciel, Sergio Santana, Andres Alvarez, Alan Collins, Gabriel Herrera, Ramon Guillen, and Carlos Soto.

(www.iteso.mx), is shown above in Photo 3. Alan Collins and Gabriel Herrera have recently graduated and formed a company (Nextia Embedded Systems) that specializes in medical systems (more specificly: embedded solutions for differently able people), including

design and development outsourcing services. For more information on the myoelectric hand or the new company, contact Alan Collins at [email protected]. Special thanks to Allie McCormick for arranging a telephone interview with Alan and Gabriel. SV

SERVO 11.2007

77

Then&Now.qxd

10/10/2007

10:17 AM

Page 78

a n d

Then NOW HUMANOID ROBOTS b

I

n my many articles over the years, I’ve managed to mention just a few tantalizing tidbits about the progress being made in humanoid robots, but have just categorized that information within the context of the particular article that I was writing. In June ’06, I wrote about the development of walking robots in this column, however, humanoid form goes quite a bit beyond simple bipedal walking machines or even the addition of a ‘head’ as the center of information-gathering sensors. The earliest robots were always anthropomorphic or ‘man formed’ and the plays and movies from the beginning of last century depicted them in this way. Of course, a human was inside a robot suit so it was much easier to use an actor to give movement to a human-sized FIGURE 1. Gort the robot and Klaatu the alien.

y

T

o

m

bipedal creature that to engineer the complex machinery required for the illusion. Gort from the classic 1951 sci-fi film The Day the Earth Stood Still (DVD available in Nuts & Volts webstore for $14.95; www.nutsvolts.com) was actually Lock Martin, a very tall doorman recruited from Grauman’s Chinese Theater in Hollywood, CA (see Figure 1). The rubber ‘Gort suit’ was so uncomfortable that Martin could only stand it for 30 minutes at a time and it was so restrictive to his movements that he could not pick up the actors that he was seen to ‘carry about.’ Wires and moveable stands were used to support the people out of sight of the camera. As an active scuba diver in the ‘60s and ‘70s, I can just imagine how uncomfortable a rubber suit could be after standing in the sun with a wet suit FIGURE 2. Maria, the robot in Metropolis.

C

a

r

r

o

l

l

on after a dive. I’m not sure if it was any more comfortable than the ‘Maria’ suit worn by actress Brigitte Helm in the earlier classic movie Metropolis, filmed in 1927 (see Figure 2). Much later, Star Wars producer George Lucas used diminutive actor, Kenny Baker crammed inside the body of R2D2 in 1977 (and all later sequels) to give life to the ‘robot,’ as mechanisms of the day could not produce the rocking motion and give movement to all of its many functions. Actor Anthony Daniels was similarly stuffed into a gold robot suit not too unlike the Maria costume of a half century earlier to give motion to C-3PO. In people’s minds decades ago, most robots had to look like bipedal humanoids and the only way to animate these robots was to place a human inside a robot suit. The cute little beeping, rolling trashcan-looking R2D2 was the exception to the humanoid appearance.

Why Humanoids? What is the draw for experimenters to build a humanoid robot rather that a wheeled machine? They are more expensive to construct, especially if the designer uses a pair of legs for propulsion instead of wheels. They are more unstable and can be easily tipped over, either accidentally or on purpose. Reliability is another factor; one bad leg joint or actuator out of many and the robot is out of commission. I’ve followed a series of postings on the Seattle Robotics Society’s website (www.seattlerobotics.org) entitled “Legs not Wheels” for over a month.

78

SERVO 11.2007

Then&Now.qxd

10/10/2007

10:17 AM

Page 79

There are strong feelings for both types of robots and the reasoning is sound for both camps. If you want to showcase your robotics talents, your biped humanoid robot walking across a table and taking a bow is certainly a great way to add a feather to your hat. Coolness comes up strong as a reason for experimenters to want to build legged robots, especially bipedal robots. They just reek of coolness. Plus, humanoids look like the creatures we are most familiar with — us. Another strong point for humanoid robots is the technical challenge required to design and build a humanoid robot. A designer of humanoids just doesn’t make two legs cycle one in front of the other. He must take into consideration the center of gravity and balance, natural resonance of the mechanisms involved, foot design, kinematics, synchronizing all of the joints and actuators, hip and upper body position, and a multitude of other factors. The builder must also take into account just where the robot will be used — outside, indoors, on a table, in a competition, or all of the above. Statically stable gaits have been the easiest for bipedal robot builders to achieve, though a few experimenters have built running robots. We must remember, however, that many large companies have investigated walking robots as a natural way for a machine to traverse a specific area, and have found out that legs are better for certain tasks, while other applications work better with wheels. General Electric built several large walking robots that ended up with no application, yet Boston Dynamics built the Big Dog robot that just might make it onto the battlefield in the near future.

Humanoid Robots Become a Reality Despite all the movies and the many decades of wishing that true humanoids were among us, it wasn’t until Honda’s series of almost human-sized robots leading up to Asimo that humanoids actually became a reality (see Figure 3). The company spent uncounted millions of dollars (billions of Yen) in research and development of this amazing robot. Certainly university labs had toyed with the idea and Marc Raibert of Carnegie-

Mellon University had developed some amazing one-legged walkers, but the very expensive Asimo gave the world’s robot experimenters the unofficial goahead to develop their own creations. Human-sized humanoid robots proved to be quite expensive to develop due to the many DOF (degrees of freedom) or axes of motion required to approximate human leg, body, and arm motions, so small humanoid walkers became very popular. Quadruped and hexapod robots had been the ‘walker of choice’ for decades among robot experimenters. The availability of numerous types of model aircraft servos as the drive source for each axis of the legs became the final solution for the potential humanoid robot designer. These servos were as cheap as $10-$15 per axis so the transition to a bipedal humanoid from the small table-top wheeled robot was a bit easier.

Humanoid Robots Become Available to Experimenters It did not take long for the major walking robot kit manufacturers to make the leap from hexapod or quadruped robots to bipedal humanoids. Mark Tilden — long known for his work with the Physics Division of the Los Alamos National Laboratory — the UK born, ex-Canadian resident is best known for his tiny BEAM (Biology, Electronics, Aesthetics, and Mechanics) robots powered by the smallest solar cells and capable of interaction with the environment without even a simple microcontroller. Quite a few of us are familiar with Isaac Asimov’s Three Laws of Robotics, but Tilden has come up with his own rules, uniquely suited to his early robot designs and their need to extract power from the environment:

FIGURE 3. Honda’s Asimo.

he worked with NASA, JPL Labs, and DARPA to develop energy efficient robots for space and other harsh environments. He formed his own company to sell his very unique creatures and then joined with WowWee Toys to sell his most noteworthy creation, the Robosapiens and its offshoot creations. Robosapiens make an ideal first robot for those interested in humanoid designs. It comes ‘ready to roll’ right out of the box and is priced as low as $100, with later improved models under $200. The cost is low, not only because of the mass production in Asia but the simplistic mechanical design that makes use of just a few motors to perform many degrees of freedom motion to the arms and legs, much the same as the old Armatron robot arm. Tilden’s humanoid robot can walk at two different speeds, turn, dance, and has two types of grippers (it reminds me of a robot lobster, in a way). It also has various sound effects (one is a caveman type of speech) and 67 pre-programmed functions. FIGURE 4. V2 Robosapiens.

1) A robot must protect its existence at all costs. 2) A robot must obtain and maintain access to its own power source. 3) A robot must continually search for better power sources. Through his work at Los Alamos, SERVO 11.2007

79

Then&Now.qxd

10/10/2007

10:18 AM

Page 80

Many Servos Required for a Humanoid

FIGURE 5. The RS Media robot from WowWee.

Robosapien can be programmed by remote control with up to 84 program steps. The second, more expensive and larger ‘V2’ (24” tall) model has a color vision system that recognizes colors and interacts with people, can track objects, and avoid obstacles (see Figure 4). It has true bi-pedal walking with multiple gaits, can lie down and get up, and has fairly articulate hands that can grasp objects with articulated fingers. The latest Robosapien RS Media is an even more advanced robot than the V2 version (see Figure 5). FIGURE 7. Futaba RBT-1.

As I mentioned earlier, the advent of the inexpensive yet unique model airplane servo made complex humanoid construction possible for the robot experimenter. Most fully articulated humanoids require 16 to 20 or more servos to approximate many of the human joint actions, both with the legs and the arms. Even at $15 per servo, this amounts to $240-$300 just for the servos. Add a microcontroller, power pack, basic sensors (visual, communications, accelerometers, etc.), and the many complex structural members and you can begin to see why many kits top $1,000. The builder can always make up for some torque deficiencies in a particular servo by applying coil springs around the axis of joints affected by gravity such as arms and legs to force the joint rotation in a direction opposite of the force of gravity. Of course, this takes a bit of practice in selecting, mounting, and loading the spring. Top-of-the-line servos (such as the Bioloid Dynamixel line) start at $44.90 for the AX-12+ that can deliver up to 222 oz. in. of torque, but these beauties are anything but a typical servo. A standard model aircraft servo with similar features will cost almost as much without the feedback to a microcontroller of angular position and velocity, and of the torque applied by the AX-12+. I’ve had several conversations with Alex Dirks of CrustCrawler about some of the features of the Dynamixel actuators. The more FIGURE 6. ROBONOVA kit.

80

SERVO 11.2007

expensive Dynamixel RX-28 and RX-64 smart actuators, costing $200 and $285 respectively, are significantly more sophisticated in their onboard intelligence and torque. For the serious humanoid designer, they are worth the money and are available at CrustCrawler (www. crustcrawler.com) and other SERVO advertisers. Humanoids can be expensive and complex to construct; it really depends on your design and budget limits.

Humanoid Kits Many experimenters opt to start from scratch, but many types of kits are available. The Hitec ROBONOVA-I was designed as a kit but is also available pre-assembled (see Figure 6). Designed for educators, students, and hobbyists, the $899 ROBONOVA-I can walk, run, do flips and cartwheels, and even dance. It can be assembled in six to eight hours with just a screwdriver. The robot derives its individual joint movements from 16 Hitec HSR-8498HB digital servos designed specifically for this robot. They feature over-voltage and current protection, have tough Karbonite gear trains, and are touted to be easily programmed with the feedback capability. The ROBONOVA-I can be modified with additional servos, optional gyros, accelerometers, speech synthesizers, and even Bluetooth and R/C transmitters and receivers. At a shade under $1,400, the Futaba RBT-1 is a bit more expensive experimenter’s humanoid robot (see Figure 7). It tips the scales at about 900 grams and is 10” tall, but has 20 small servos, 11 of which are the Futaba RS301CR high torque metal gear servos for the leg movements, and nine lightweight RS302CD servos for the upper body. A three-axis accelerometer system allows the robot to sense motions, falling, etc., and it knows when to pick itself up after a fall. The robot’s RPU-11 controller contains an ATmega 128 CPU that interfaces with the various robot systems with an RS-485 internal link and an RS-232C interface to talk with an external Windows PC, or through a 2.4 GHz wireless controller. This little guy has a lot of degrees of freedom and should offer the serious experimenter a sophisticated research platform. The third humanoid robot kit that

Then&Now.qxd

10/10/2007

10:19 AM

Page 81

I’d like to discuss is the $899 Robotis Bioloid Comprehensive Kit that can be used to construct over 26 robots, including a pretty sophisticated humanoid robot (see Figure 8). What makes this kit so appealing to me is that the kit contains 18 of the Dynamixel AX-12+ smart actuators that total out to over $800 alone if purchased separately. This kit is great for learning humanoid dynamics as the included software, controller, AX-S1 sensor module, battery, and charger, plus the many types of quality industrial plastic structural components allow one to make some amazing things.

David Hanson’s Humanoids

FIGURE 8. Robotis Bioloid humanoid.

features and subtle muscle movements I would be remiss if I didn’t menare extremely difficult to produce, but tion the spectacular lifelike humanoid this is what draws Hanson to the task. robots created by David Hanson. David The Einstein robot head is actuated Hanson has produced a unique style of by 33 servo motors and related linkages, humanoid that very few other robot and requires just 10W of power at 6V to builders have yet to do. The faces of his achieve its full range of expressions. robots mimic the appearance and movePeople have compared his creations to the ments of a real human face. Outside of many animatronic displays at Disneyland a few robotic mannequins produced in and other theme parks. They require many Asia, virtually all humanoid robots still kilowatts of electric power and even have abstract faces and expressions. hidden hydraulic and pneumatic power People have found that realistic sources to operate. The Abraham Lincoln human facial features rendered on exhibit is a great example. Created robotic faces are pretty much creepy. decades ago, it was quite lifelike but had “That’s a view I completely reject. We many motors, linkages, and power are naturally attracted to faces and sources under the stage which were fed gestures,” says Hanson, president of through Lincoln’s feet (see Figure 11). Hanson Robotics, Inc. “Robots don’t Hanson is convinced that the major just have to make the right expression, contributor to the success of his robots is they have to make the right expression the special skin that he has developed. His at precisely the right time,” he says. ‘Frubber’ — a patented silicone elastomer Hanson began his interest in realistic whose mechanical properties allow the robot faces when he studied art at the complex facial movements — is a foamed Rhode Island School of Design and later platinum-based elastomer that contains received a Ph.D. in Interactive Arts and up to 70 percent air by volume. The Engineering from the University of Texas. control of the size and distribution of the His well-known renderings of Albert Einstein’s head applied to a FIGURE 11. Abraham Lincoln animatronic figure. Korean-designed humanoid robot is a bit on the creepy side (see Figure 9). I found his rendering of Sci-Fi writer, Philip K. Dick (see Figure 10) absolutely amazing as it displays many thousands of nuanced, believable facial expressions. Many people steer clear of very realistic human faces on humanoid robots as the facial

FIGURE 9. Einstein-Hubo combo.

open and closed air cells in the Frubber skin is what allows it to move much like human skin. The fact that it can be moved by small servos with little FIGURE 10. force makes it use- Hanson’s Philip K. Dick robot head. ful for humanoid robot faces. It is so flexible that it can be stretched as much as 900%, though damage may occur beyond 450% stretching. Figure 12 illustrates how Hanson has mimicked human facial expressions with his Frubber-based robot faces. Hanson is working with well-known Japanese robot designer, Tomotaka Takahashi to produce RoboKind, a robot FIGURE 12. Photo by Chris Buck of Hanson’s robot’s facial expressions.

SERVO 11.2007

81

Then&Now.qxd

10/10/2007

10:20 AM

Page 82

that will be 14 inches tall with a body designed by Takahashi and a head designed by Hanson. These cartoon-like robots will be able to walk around and offer a range of facial expressions. “Biped robots aren’t that unusual in Japan. There are even soccer matches for them,” says Hanson. “But our new robot will be the only one capable of complex facial expressions.” A limited edition version of the new robot will sell for about $10,000

and a standard model will cost about $3,000. Hanson is also studying many other more efficient power sources for the appendages and facial movements of his humanoids. He feels that humanoids are the ideal robot design to interact with people. As with all my articles, I have only scratched the surface of the many technologies that I discuss. The vast potential of humanoid robots is just beginning. It is you — the readers of this magazine — who

will forge those steps to finally produce that one humanoid who will be indistinguishable from a human being. Maybe not in our lifetime, but there will be humanoids that can stand beside us and be the robotic equivalent of Turing’s Test for AI. Meanwhile, go to the various manufacturers such as CrustCrawler, Trossen Robotics, Lynxmotion, or Robotis to see what is already available. Buy a kit or ready-made robot, modify it, and then create your own. Take that giant step. SV

Advertiser Index Active Robots .............................................3 All Electronics Corp. ..........................31, 46 AP Circuits/e-pcb.com ............................60 AWIT ..........................................................46 CrustCrawler .............................................13 Electronics123 ..........................................31 Floatation Center — Art Gallery ..............77 Futurlec .....................................................46 Gears Educational Systems, LLC .............61 Hitec ..........................................................16 Hobby Engineering .................................46

82

SERVO 11.2007

Images Co. ................................................46 IMService ............................................31, 46 Jameco .................................................2, 46 Lorax Works ........................................31, 46 Lynxmotion, Inc. .......................................19 Maxbotix ...................................................46 Maximum Robotics ............................20, 46 Net Media .................................................83 Parallax, Inc. ...............................Back Cover PCB Pool .............................................46, 55 Pololu Robotics & Electronics ..........21, 46 Robotis Co. Ltd. ..........................................7

Robot Power ............................................66 RobotShop, Inc. .................................46, 82 Schmartboard .....................................31, 75 SCON .........................................................31 Solarbotics/HVW .......................................9 SORC ..........................................................60 SPSU ...........................................................17 Technological Arts ...................................46 TORMACH .................................................12 Vantec .......................................................66 Yost Engineering .......................................71

CoverInside.qxd

10/10/2007

11:34 AM

Page 2

We’re passive aggressive When it comes to passive products, we don’t pull any punches: we stock more major brands of passive components than any other major catalog distributor.* So whatever brands you need—

5

10

15

20

from AMP or AVX to Vishay or Wakefield— you’re more likely to find them all at Jameco. Check our “stats” below and see for yourself. It’s another Jameco advantage.

Jameco®

Digi-Key®

Mouser®

Newark®

AMP Astec Power Augat AVX Bourns Cherry Condor Power Corcom CTS Elco Connector Fox Electronics Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Power-One Teledyne Relays Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Corcom CTS Elco Connector Fox Electronics Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Power-One Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Condor Power Corcom Elco Connector Fox Electronics Int’l Rectifier Kemet Molex Panasonic Potter & Brumfield Teledyne Relays Vishay Wakefield

AMP Astec Power AVX Bourns Cherry Corcom Elco Connector Grayhill Int’l Rectifier Kemet Molex Osram Panasonic Potter & Brumfield Vishay Wakefield

Free shipping on these and 83 other major brands. Call for details.

OTHER JAMECO ADVANTAGES:
More major brands of semis than any other catalog.
99% of catalog products ship the same day.
Lowest prices guaranteed, or we pay 10%.
Major brand names and generic equivalents

for even greater cost savings.

Order 24 hours a day, 7 days a week

www.Jameco.com Or call 800-831-4242 anytime ©Jameco Electronics. *According to the line cards on their web sites on August 28, 2007. Trademarks are the property of their respective owners.

Page 84 7:56 PM 10/10/2007 Cover.qxd

Vol. 5 No. 11

SERVO MAGAZINE

TOUCH BIONICS • KILLER ROBOTS • HUMANOIDS • ANDROID ARMS

November 2007

4

74470 58285 0

1 1>

$7.00 CANADA

$5.50 U.S.