Detection And Tracking Of Moving Object With A Mobile Robot Using Laser Scanner

DETECTION AND TRACKING OF MOVING OBJECT WITH A MOBILE ROBOT USING LASER SCANNER JIN-XIA YU , Henan Polytechnic University ZI-XING CAI , Central South University ZHUO-HUA DUAN , Shaoguan University

Proceedings of the Seventh International Conference on Machine Learning and Cybernetics, Kunming, 12-15 July 2008

In This Paper, An autonomous approach for detection and

tracking of moving object with a mobile robot using laser scanner is presented Firstly, ranging data of environmental objects from laser scanner are clustered Secondly, the movement parameter of clustering objects is computed by local grid map matching. After obtaining the moving object, particle filter (PF) with the improved proposal distribution is adopted to track moving object

Introduction Biswas and colleagues proposed an algorithm

called Dogma Wang presented a motion-based detector to detect different kinds of moving objects and a hypothesis tree to manage data association Yang investigated the time-varying potential field algorithm that is favorable to realize path planning and objects avoidance in mobile robot navigation

Introduction Schulz introduced a sample-based variant of

joint probabilistic data association filters to track features originating from individual objects and to solve the correspondence problem between the detected features and the filters Recently, laser scanner has been widely applied into the field of detection and tracking -high ranging precision -good real-time performance

Detection and tracking of Grid map building -The objects information is the snapshot from the laser scanner LMS291. -Detected Points are represented in Polar Coordinate System -Operating environment of mobile robot is described by 2D Cartesian grids

Detection and tracking of moving object Spatial clustering of objects

(1) Dynamic clustering of laser ranging data -Difference between measured points

-If < threshould , objects between is considered to be in the same region -Otherwise, belongs to a new region -New object region is grouped repeating the procedure until j reaches 360; so 361 laser measurements can be divided into c different object regions.

Detection and tracking of moving object 

Spatial clustering of objects (2) Characteristics of clustering objects -Centroid of object region

Detection and tracking of moving object 

Spatial clustering of objects (2) Characteristics of clustering objects -Moving Velocity of object region

-The quantity of object region is defined as the grid number of its region and its centroid

Detection and tracking of moving object  Moving object detection based on grid map matching *Algorithm:

-Step 1: to read the real-time environment information in detection window from laser scanner at current sample period, and to build and save the grid map of the detection window. -Step 2: to read the same information at next sample period, and to build the linked list of objects at T+1 period as the method shown in step 1 -Step 3: to search the linked list at T+1 periods, and to match with that at T periods. The matching criterion is the distance of the estimated coordinates between two objects is below the threshold

Detection and tracking of moving object  Moving object detection based on grid map matching *Algorithm:

-Step 4: to implement the local grid map matching at T and T+1 periods so as to gain the motion parameter of the same object such as the moving distance and orientation -Step 5: to determinate the flag bit, that is the motion condition of object (static or dynamic), by judging the moving distance dΔ is below the threshold δ whether or not. Then this record is inserted the linked list of object -Step 6: for the same object, to evaluate the linked list and the grid map at T periods to at T+1 periods,

Detection and tracking of moving object Movement compensation of mobile robot -Measurement Errors

: measurement error of mobile robot compared with moving object : denotes the ranging error of laser scanner compared with moving object that can be determined : denotes the relative motion error between mobile robot and moving object

Detection and tracking of moving object Moving object tracking based on the

improved particle filter -In tracking process, EKF is usually adopted to estimate moving object; but it exists bigger truncation errors -In recent, particle filter has been paid close attention for its powerful on-line estimation

Detection and tracking of moving object -Motion and observation equation of moving object

denotes the position V is process noise relative position of mobile robot against moving object robots position and orientation W is observation noise

Estimation and Update Phase of Moving Object with PF

Square-Root Unscented Kalman Filter (SRUKF)

Square-Root Unscented Kalman Filter (SRUKF)

Experimental Results

Experimental Results

Experimental Results