Optimizing ASC/3 Signal Timings through SILS - Traffic Signal ...
Optimizing ASC/3 Signal Timings through SILS - Traffic Signal ...
Optimizing ASC/3 Signal Timings through SILS - Traffic Signal ...
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
<strong>Optimizing</strong> <strong>ASC</strong>/3 <strong>Signal</strong><br />
<strong>Timings</strong> <strong>through</strong> <strong>SILS</strong><br />
Aleksandar Stevanovic<br />
<strong>Traffic</strong> <strong>Signal</strong> Systems Committee<br />
Summer Meeting 2008
GA in <strong>Traffic</strong> Control Optimization<br />
Pros:<br />
Use micro simulation for evaluation – better<br />
representation of traffic<br />
Optimization in complex search space<br />
Cons:<br />
Long optimization time<br />
No guaranties for global optimum<br />
Encoding and Fitness Function should be<br />
carefully chosen<br />
2
VISGAOST<br />
Practical tool for the optimization of signal timings<br />
Uses<br />
VISSIM<br />
Simulates l t traffic<br />
Provides evaluations of Measures of Effectiveness (MOE)<br />
Genetic Algorithms (GA)<br />
Evolve generations of timing plans mimicking natural<br />
evolution<br />
Find better timing gplans based on provided VISSIM’s<br />
evaluations
VISSIM-based GA Optimization of<br />
<strong>Signal</strong> <strong>Timings</strong> i (VISGAOST)<br />
no<br />
yes<br />
4
GA Optimization Inputs<br />
5
Distributed GA Processing<br />
Example:<br />
6 Timing Plans<br />
4Computers<br />
6
Park City, UT
Results – Park City, UT
Albany, NY
Results – Albany, NY
Detroit, MI<br />
Woodward Ave.<br />
13 miles long<br />
70 actuated<br />
intersections<br />
Source: Transit <strong>Signal</strong> Priority Evaluation Results Woodward Avenue. PTV America.<br />
11
Results – Detroit, MI<br />
9000<br />
Average Person Delay<br />
8750<br />
Best Person Delay<br />
8500<br />
Pe erson Delay (sec)<br />
8250<br />
8000<br />
7750<br />
7500<br />
0 200 400 600 800 1000 1200 1400 1600 1800 2000 2200 2400 2600 2800 3000<br />
Number of Evaluations<br />
12
<strong>Traffic</strong> Control in Microsimulation<br />
Emulation-In-the-Loop Simulation (EILS)<br />
Application Programming Interface (API) used<br />
to model traffic control<br />
Hardware-In-the-Loop Simulation (HILS)<br />
Actual hardware connected to traffic<br />
simulation<br />
Software-In-the-Loop Simulation (<strong>SILS</strong>)<br />
Actual software (installed on hardware) used<br />
to control traffic in simulation models
EILS – VISSIM & NEMA Emulator<br />
<strong>Traffic</strong> Simulator - VISSIM<br />
<strong>Traffic</strong> Control Emulator<br />
Based on NEMA standards
Peek LMD 9200 Emulator
Software-In-the-Loop Simulation<br />
Advantages over HILS<br />
Enabling faster than real-time simulator-controller<br />
operations<br />
No traffic control hardware<br />
Synchronization h i of controller’s and computer’s clocks<br />
The first noncommercial <strong>SILS</strong> - Siemens ITS<br />
(2002) - NextPhase & CORSIM and VISSIM<br />
The first commercial <strong>SILS</strong> - VISSIM & <strong>ASC</strong>/3
<strong>SILS</strong> – VISSIM & <strong>ASC</strong>/3<br />
<strong>Traffic</strong> Simulator - VISSIM<br />
<strong>Traffic</strong> Controller<br />
(Software)<br />
As installed in field
<strong>SILS</strong>-VISGAOST Connection<br />
<strong>SILS</strong><br />
<strong>ASC</strong>/3 Software<br />
VISSIM<br />
<strong>ASC</strong>/3 Database File<br />
ª ¯¬«¼®¨€@PQ<br />
"CBRƒcq‚sbar<br />
ˆ‰®¤¢¨€@P Taq<br />
% \ - Ç Æ Á Â Ð ¹ º<br />
Ñ P ! Q A 1"‚ - @ \ aA<br />
0`€@PQ " # C Sc 2<br />
B R 3CS p €bBR ƒ c<br />
` q ‚ s b a rˆ‰ W ' & % $<br />
D TTDŠ 7 4d<br />
G V U EeEUgGW 6 F<br />
5fFV¤µÅ-¢<br />
Æ¡³£®´®´°<br />
±®·¥®®Ä'øñr<br />
úrrŽr!¢ T dŒ ‹ V é À<br />
VISSIM Output File<br />
Network<br />
Performance<br />
VISGAOST<br />
<strong>ASC</strong>/3 Interface<br />
A 01000001 65<br />
& 00100110 38<br />
3 00110011 51<br />
# 00100011 35<br />
• 10010101 149<br />
“ 10010011 147<br />
GA - Optimizer<br />
Simulation time from 600 to 4200<br />
Parameter Value<br />
Total travel time[h],All Veh 835.8<br />
Total delay time[h],All Veh 159.2<br />
Average speed[mph],All Veh 28.9<br />
Number of stops,All Veh 21828<br />
Distance travelled,All Veh 24189<br />
Stopped delay[h],All Veh 84.2
Case Study – 3500 South, SLC<br />
5200 W<br />
4800 W<br />
4400 W<br />
4155 W<br />
4000 W
Calibration & Validation<br />
1800<br />
s [veh/h]<br />
Simu ulation traffic count<br />
1600<br />
1400<br />
1200<br />
1000<br />
800<br />
600<br />
400<br />
R² = 0.988<br />
200<br />
0<br />
0 200 400 600 800 1000 1200 1400<br />
Field traffic counts [veh/h]<br />
Mo odel travel times pe er section [sec]<br />
75<br />
65<br />
55<br />
45<br />
35<br />
R² = 0.986<br />
25<br />
25 35 45 55 65 75<br />
Field travel times per section [sec]
Optimization of PI<br />
Perfor rmance Index<br />
30<br />
25<br />
20<br />
Initial Performance Index<br />
Average Performance Index<br />
Best Performance Index<br />
15<br />
10<br />
Cycle Length<br />
Cycle Length<br />
&<br />
Offsets<br />
Cycle Length,<br />
Offsets &<br />
Phase Sequence<br />
0 50 100 150 200 250<br />
Splits<br />
Number of Generations
Thank you<br />
Questions