Traffic Light Control
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View Traffic Light Control as PDF for free.
More details
- Words: 1,058
- Pages: 41
Traffic Light Control
Hoàng Hải Lưu Như Hòa Department of Automatic Control Hanoi University of Technology
11/19/08
1
Overview The problem of transport system is an optimal problem control Main Goals are:
Improving safety Minimizing travel time Increasing the capacity of infrastructures
11/19/08
2
Outline
Section 1: How traffic can be modeled, predicted and controlled ? Section 2: What a traffic light control system contain ? Section 3: New approaches to traffic light control !!!
11/19/08
3
Modeling and Controlling Traffic Section 1
11/19/08
4
Modeling and Controlling Traffic How traffic can be modeled ?
Macroscopic scale: Similar to models of fluid dynamics PDE
11/19/08
Microscopic scale: each vehicle is considered as an individual ODE
Modeling and Controlling Traffic
5
Macroscopic models
Macroscopic models based on fluid dynamics model Relation between: traffic flux, traffic density and velocity.
11/19/08
Modeling and Controlling Traffic
6
Macroscopic models
11/19/08
Modeling and Controlling Traffic
7
Macroscopic models Basic Statements The more vehicles are on a road, the slower their velocity will be The number of vehicles entering the control zone has to be smaller or equal to the number of vehicles leaving the zone in the same time At a critical traffic density and a corresponding critical velocity the state of flow will change from stable to unstable. If one of the vehicles brakes in unstable flow regime the flow will collapse 11/19/08
Road Capability, Speed Limit
not simulate directly certain driver behaviors
Modeling and Controlling Traffic
8
Microscopic models
In contrast to macroscopic models, microscopic models focus on vehicles (position and velocity ) Cellular automaton (CA): discrete model
Stephen Wolfram Creator of CA
Road Δx Time steps Δt Nagel-Schreckenberg model
11/19/08
Modeling and Controlling Traffic
9
Microscopic models Road Cell
Rule
18410 101110002
current pattern new state for center cell
11/19/08
If next state is available Then Move forwards Else Stop
111 110 101 100 011 010 001 000 1 0 1 1 1 0 0 0 Modeling and Controlling Traffic
10
Microscopic models
Self-caused slowdown:
Stable "stop-waves“
Two stable states
11/19/08
Modeling and Controlling Traffic
11
Microscopic models
Cognitive Multi-Agent Systems (CMAS): Agents interact and communicate with each other and the infrastructure
receives information from the environment using its sensors believes certain things about its environment uses these beliefs and inputs to select an action using learning capabilities to optimize agent
11/19/08
Modeling and Controlling Traffic
12
Microscopic models
11/19/08
Modeling and Controlling Traffic
13
Microscopic models Dia (2002) use CMAS in traffic problem. But no result were presented!!!
11/19/08
Modeling and Controlling Traffic
14
Predicting Traffic
Measuring traffic over a certain time, assuming that conditions will be the same for the next period Ledoux(1996) used neural networks perform long-term prediction of the queue length at a traffic light
11/19/08
Modeling and Controlling Traffic
15
Vehicle Control
Get information through dynamic road signs, radio, or even on-board navigation systems Traffic flow increase if all drivers drive at the same (maximum) speed. (But …) Learned strategies better than handcrafted controllers
11/19/08
Modeling and Controlling Traffic
16
Traffic Light Control System Section 2
11/19/08
17
Traffic Light Control System
Distributed System
A set of intersections A set of connection (roads) Traffic lights regulating Traffic lights are controlled independently
11/19/08
Traffic Light Control System
18
Traffic Control and Command Centre In Thailand
11/19/08
Traffic Light Control System
19
Traffic Light Control System
No obvious optimal solution In practice most traffic lights are controlled by fixed-cycle controllers Fixed controllers need manual changes to adapted specific situation
11/19/08
Traffic Light Control System
20
Green Waves
Offset of cycle can be adjusted to create green waves.
11/19/08
Traffic Light Control System
21
Driver Detector - Sonar Sensor
•Few drivers •Unusual
11/19/08
Traffic Light Control System
22
Driver Detector - Camera
Identification image Expensive Complex Traffic System
11/19/08
Traffic Light Control System
23
11/19/08
24
Driver Detector - Loop Detector •Measure Inductive •Most popular •Cheap
11/19/08
Traffic Light Control System
25
Traffic Light Control System What does it do ?
11/19/08
Traffic Light Control System
26
Let’s See!
N
W
S 11/19/08
27
N
No turning
W
S 11/19/08
28
Binary traffic lights
N
W
11/19/08
S
29
Safety Property
N
W This should not happen 11/19/08
S
30
Safety Propert y
N
W
This should not happen 11/19/08
S
31
Livenes s Propert y
N
W Traffic in each direction must be
11/19/08
Thank God!
S 32
Finite State Machine
The Problem is Synchronous
11/19/08
Traffic Light Control System
33
Finite State Machine
11/19/08
Traffic Light Control System
34
Control Algorithms
Section 3
11/19/08
35
Expert Systems
uses a set of given rules to decide upon the next action (change some of the control parameters) Findler,Stapp,1992 describe a network of roads connected by traffic lightbased expert systems improve performance but much computation
11/19/08
Control Algorithms
Can Machines Think?
36
Evolutionary Algorithms
Taaleetal,1998 using evolutionary algorithms to evolve a traffic light controller for a single intersection Result: Generates green times for next switching schedule. Minimization of total delay / number of stops. Better results (3 – 5%) / higher flexibility than with traditional controllers. Dynamic optimization, depending on actual traffic (measured by control loops).
11/19/08
Control Algorithms
37
Fuzzy Logic
Passed through 31% more cars Average waiting time shorter by 5% Performance also measure 72% higher. In comparison with a human expert the fuzzy controller passed through 14% more cars with 14% shorter waiting time and 36% higher performance index
11/19/08
Control Algorithms
38
Reinforcement Learning
Reinforcement learning is a branch of machine learning concerned with how an agent ought to take actions in an environment so as to maximize some notion of long-term reward Thorpe used a neural network for the traffic-light based value function which predicts the waiting time for all cars standing at the junction
11/19/08
Control Algorithms
39
Intelligent Agents
11/19/08
Control Algorithms
40
References
http://en.wikipedia.org [Wiering] Intelligent Traffic Light Control [Tan Kok Khiang] Intelligent Traffic Lights Control by Fuzzy Logic
11/19/08
41
Hoàng Hải Lưu Như Hòa Department of Automatic Control Hanoi University of Technology
11/19/08
1
Overview The problem of transport system is an optimal problem control Main Goals are:
Improving safety Minimizing travel time Increasing the capacity of infrastructures
11/19/08
2
Outline
Section 1: How traffic can be modeled, predicted and controlled ? Section 2: What a traffic light control system contain ? Section 3: New approaches to traffic light control !!!
11/19/08
3
Modeling and Controlling Traffic Section 1
11/19/08
4
Modeling and Controlling Traffic How traffic can be modeled ?
Macroscopic scale: Similar to models of fluid dynamics PDE
11/19/08
Microscopic scale: each vehicle is considered as an individual ODE
Modeling and Controlling Traffic
5
Macroscopic models
Macroscopic models based on fluid dynamics model Relation between: traffic flux, traffic density and velocity.
11/19/08
Modeling and Controlling Traffic
6
Macroscopic models
11/19/08
Modeling and Controlling Traffic
7
Macroscopic models Basic Statements The more vehicles are on a road, the slower their velocity will be The number of vehicles entering the control zone has to be smaller or equal to the number of vehicles leaving the zone in the same time At a critical traffic density and a corresponding critical velocity the state of flow will change from stable to unstable. If one of the vehicles brakes in unstable flow regime the flow will collapse 11/19/08
Road Capability, Speed Limit
not simulate directly certain driver behaviors
Modeling and Controlling Traffic
8
Microscopic models
In contrast to macroscopic models, microscopic models focus on vehicles (position and velocity ) Cellular automaton (CA): discrete model
Stephen Wolfram Creator of CA
Road Δx Time steps Δt Nagel-Schreckenberg model
11/19/08
Modeling and Controlling Traffic
9
Microscopic models Road Cell
Rule
18410 101110002
current pattern new state for center cell
11/19/08
If next state is available Then Move forwards Else Stop
111 110 101 100 011 010 001 000 1 0 1 1 1 0 0 0 Modeling and Controlling Traffic
10
Microscopic models
Self-caused slowdown:
Stable "stop-waves“
Two stable states
11/19/08
Modeling and Controlling Traffic
11
Microscopic models
Cognitive Multi-Agent Systems (CMAS): Agents interact and communicate with each other and the infrastructure
receives information from the environment using its sensors believes certain things about its environment uses these beliefs and inputs to select an action using learning capabilities to optimize agent
11/19/08
Modeling and Controlling Traffic
12
Microscopic models
11/19/08
Modeling and Controlling Traffic
13
Microscopic models Dia (2002) use CMAS in traffic problem. But no result were presented!!!
11/19/08
Modeling and Controlling Traffic
14
Predicting Traffic
Measuring traffic over a certain time, assuming that conditions will be the same for the next period Ledoux(1996) used neural networks perform long-term prediction of the queue length at a traffic light
11/19/08
Modeling and Controlling Traffic
15
Vehicle Control
Get information through dynamic road signs, radio, or even on-board navigation systems Traffic flow increase if all drivers drive at the same (maximum) speed. (But …) Learned strategies better than handcrafted controllers
11/19/08
Modeling and Controlling Traffic
16
Traffic Light Control System Section 2
11/19/08
17
Traffic Light Control System
Distributed System
A set of intersections A set of connection (roads) Traffic lights regulating Traffic lights are controlled independently
11/19/08
Traffic Light Control System
18
Traffic Control and Command Centre In Thailand
11/19/08
Traffic Light Control System
19
Traffic Light Control System
No obvious optimal solution In practice most traffic lights are controlled by fixed-cycle controllers Fixed controllers need manual changes to adapted specific situation
11/19/08
Traffic Light Control System
20
Green Waves
Offset of cycle can be adjusted to create green waves.
11/19/08
Traffic Light Control System
21
Driver Detector - Sonar Sensor
•Few drivers •Unusual
11/19/08
Traffic Light Control System
22
Driver Detector - Camera
Identification image Expensive Complex Traffic System
11/19/08
Traffic Light Control System
23
11/19/08
24
Driver Detector - Loop Detector •Measure Inductive •Most popular •Cheap
11/19/08
Traffic Light Control System
25
Traffic Light Control System What does it do ?
11/19/08
Traffic Light Control System
26
Let’s See!
N
W
S 11/19/08
27
N
No turning
W
S 11/19/08
28
Binary traffic lights
N
W
11/19/08
S
29
Safety Property
N
W This should not happen 11/19/08
S
30
Safety Propert y
N
W
This should not happen 11/19/08
S
31
Livenes s Propert y
N
W Traffic in each direction must be
11/19/08
Thank God!
S 32
Finite State Machine
The Problem is Synchronous
11/19/08
Traffic Light Control System
33
Finite State Machine
11/19/08
Traffic Light Control System
34
Control Algorithms
Section 3
11/19/08
35
Expert Systems
uses a set of given rules to decide upon the next action (change some of the control parameters) Findler,Stapp,1992 describe a network of roads connected by traffic lightbased expert systems improve performance but much computation
11/19/08
Control Algorithms
Can Machines Think?
36
Evolutionary Algorithms
Taaleetal,1998 using evolutionary algorithms to evolve a traffic light controller for a single intersection Result: Generates green times for next switching schedule. Minimization of total delay / number of stops. Better results (3 – 5%) / higher flexibility than with traditional controllers. Dynamic optimization, depending on actual traffic (measured by control loops).
11/19/08
Control Algorithms
37
Fuzzy Logic
Passed through 31% more cars Average waiting time shorter by 5% Performance also measure 72% higher. In comparison with a human expert the fuzzy controller passed through 14% more cars with 14% shorter waiting time and 36% higher performance index
11/19/08
Control Algorithms
38
Reinforcement Learning
Reinforcement learning is a branch of machine learning concerned with how an agent ought to take actions in an environment so as to maximize some notion of long-term reward Thorpe used a neural network for the traffic-light based value function which predicts the waiting time for all cars standing at the junction
11/19/08
Control Algorithms
39
Intelligent Agents
11/19/08
Control Algorithms
40
References
http://en.wikipedia.org [Wiering] Intelligent Traffic Light Control [Tan Kok Khiang] Intelligent Traffic Lights Control by Fuzzy Logic
11/19/08
41
Related Documents
Intelligent Traffic Light Control
November 2019 8
Traffic Light Control
November 2019 4
Traffic Light Control
November 2019 2
Traffic Control
July 2020 3
Air Traffic Control
November 2019 16
Air Traffic Control
October 2019 7More Documents from ""
Identification Toolbox
June 2020 0
Traffic Light Control
November 2019 2