55
Adaptive signal Control Tom Mathew
Adaptive signal Control
Tom Mathew
Adaptive Control: Outline
1. Signal Control Taxonomy
2. Coordinated Signal System
3. Vehicle Actuated System
4. Area Traffic Control (Responsive)
5. Adaptive Traffic Control Systems (1, II, III)
2
I. Signal Control Taxonomy
Traffic Signal Control
Fixed Time Signal Vehicle Actuated
Coordinated Signal Area Traffic Control
Responsive
4
Adaptive
Coordinated control
Offset downstream
green start
Assumes constant link
travel time
Implies smooth
progression
5
Traffic Signal Control
Traffic Signal Control
Fixed time control
6
Vehicle Actuated Control
Vehicle Actuated Control Concept
8
Vehicle Actuated Control Working principle
9
Vehicle Actuated Control
Parameters
Minimum Green time
Maximum Green time
Threshold gap
Unit extension
10
Vehicle Actuated ControlBasic Algorithm
for every phase
set green equals queue service time
for every scan time
get detector state for each lane-group
compute gap
if gap greater than threshold
terminate green
else increment green time
limit green to max green time
if green greater than max green
terminate green
11
Traffic Signal Control
Area traffic control – Traffic responsive
13
Traffic Signal Control
Area traffic control – Adaptive to Traffic
14
Area (Adaptive) Traffic Control
Traffic Signal Control
Two Popular Network Systems
Centralized system
SCOOT
Split, Cycle, Offset, Optimization
Distributed system
SCAT
Sydney Coordinated Adaptive Traffic System
16
SCOOT system
Working philosophy
Upstream detection
Data communicated to
central controller
It computes the timing and
send to intersections
Limitations
Communication overheads
Poor progression prediction
Calibration issues
17
SCATS system
Working philosophy
Downstream detection
Local controller acts
akin to a VA controller
Communicate
periodically to the
central controller
Limitations
Not an optimal system
18
SCOOT vs. SCAT SCOOT
Centralized System
Upstream detection
Fixed traffic regions
Fallback - fixed
Modal based
SCAT
Distributed system
Stop line detection
Adjustable region
Fallback - VA
Heuristic
19
Adaptive Control – I
CoSiCoSt
Composite Signal Control Strategy
Developed by: CDAC(T)
Features
Use the philosophy of Split, Cycle and Offset
approximation for optimum throughput
Every controlled lane require detection
Use stop-line and exit detectors
Built-in filters to segregate turning traffic
21
CoSiCoSt
CoSiCoSt
22
Detector placement
Stop line - No demand
prediction
Input
Demand from every loop
from every cycle
Output
Green time for each
phase, Cycle length and
delay
CoSiCoSt
23
Heuristic algorithm
Based on actual &
utilized green time
CoSiCoStBasic Algorithm of VA
for every phase
set green equals queue service time
for every scan time
get detector state for each lane-group
compute gap
if gap greater than threshold
terminate green
else increment green time
limit green to max green time
if green greater than max green
terminate green
24
CoSiCoSt Working principle of VA
25
• Indian Adaptive Traffic Control System
• Reactive – Rule based adjustments
• Distributed network architecture
• Real-time Signal Coordination of dynamically
selected corridors
• Self-calibrating Cycle lengths and Phase lengths
• Fall-back operation in VA
CoSiCoSt
• Light weight
– Operates on standard desktop
– Windows / Linux Platform
– PostgreSQL database
– TCP/IP Protocol
– Optimized data transfer
– Networking possible even on Broadband VPN
• Scalable from two junctions to 6000 junctions
– 200 corridors x 30 junctions on a 32-bit machine
CoSiCoSt
Central Server
• Web Enabled
• Data Logger
• Configurator
• User friendly GUI
• Remote Monitoring
• Reports, Views and Graphs
• ATCS Application Software
• Real-time animation of Signal Operation
• Real-time animation of Time-Space Diagram
CoSiCoSt
Implementation
• Jaipur, Rajasthan
• Pune, Maharashtra
• Kolkata, West Bengal
• Ahmedabad, Gujarat
• Patna, Bihar (in progress)
• Surat, Gujarat (in progress)
CoSiCoSt
19 April 2018
Dynamic Offset
• Progression modeling
• Improve offset
calculation
• Establish relation
between link travel
time, loop occupancy
and link length
30
CoSiCoSt
Adaptive Control – II
Machine Learning
Adaptive controlAdaptive learning – Estimate of Gmax
32
Curr
ent
Gre
en T
imes
Actual discharges Max Green
Reward rtState st Action at
Agent
Environment
rt+1
St+1
Adaptive controlEvaluation using traffic simulator
33
Adaptive controlResults
34
Volume Control Delay Queue
Low
(V/C 0.3-0.5)
Adaptive 15 6
VA 17 6
Fixed 20 8
Medium
(V/C 0.5-0.8)
Adaptive 20 11
VA 23 14
Fixed 28 20
High
(V/C 0.8-1.2)
Adaptive 44 66
VA 53 93
Fixed 67 129
0
5
10
15
20
25
30
35
40
45
50
0 5000 10000 15000 20000 25000 30000 35000 40000
Gre
en
(s)
Time (s)
Green Time by VA control
Green-1
0
10
20
30
40
50
60
0 5000 10000 15000 20000 25000 30000 35000 40000
Vo
lum
e (
veh
)
Time (s)
Volume
Vol-1
0
10
20
30
40
50
60
70
80
90
100
0 5000 10000 15000 20000 25000 30000 35000 40000
Gre
en
(s)
Time (s)
Green Time by Adaptive control
Green-1
Adaptive control
Adaptive Control – III
Optimization of Modified HCM
Adaptive control Objective
Minimize average delay per vehicle
Decision variable
Green time for each phase
Constraints
Upper and lower limits for green times
Maximum cycle length
Input
Not upstream volume, but stop line discharge37
Adaptive control Mathematical formulation
38
Adaptive control Delay function (HCM 2000)
39
Adaptive control Delay function (HCM 2000)
40
qp
Evaluation Numerical example
41
Evaluation
42
Evaluation
43
Evaluation Input demand
44
Evaluation Output - Cycle
45
Evaluation Output – Green times
46
47
Evaluation Output – Delays
48
Evaluation (Smoothening)
49
Adaptive Control
Summary
Sensitive to fluctuating traffic demand
Evaluation by traffic simulators
Optimal use of infrastructure
Enhances service quality
50
Adaptive Control
Advanced topics
Developing for large systems
Better traffic model
Heuristics
Machine Learning
Delay Models
Traffic management capabilities51
ATMS Detection system
Vehicle
Decision
Optimal
Control system
Intersection control
System optimal
52
53
Adaptive Control
• Summary
–Sensitive to fluctuating traffic demand
–Evaluation by traffic simulators
–Optimal use of infrastructure
–Enhances service quality
57
Adaptive Control
• Research directions
–Developing for large systems
–Optimal control
–Traffic management capabilities
58
