Game-Theory Based Traffic Signal Control Algorithm Using Drone-based and Simulated Traffic Data Reduced Traffic Delay by 37.7 Percent and Turning Movement Estimation Error by Up To 50 Percent.
Adaptive Signal Control Algorithm with Connected Vehicles Tested Using Data from Toronto and Orlando.
Orlanda, Florida, United States
Toronto, Ontario, Canada
Integration of a Real-time Traffic State Estimation and a Decentralized Game-Theoretic Traffic Signal Controller
Summary Information
Real-time adaptive traffic signal control is an ITS solution designed to optimize traffic flow at intersections, reduce congestion, and minimize delays. Traditional formulas to optimize cycle length rely on static assumptions, but advancements in connected vehicle (CV) technology provide an opportunity to enhance signal control via improved traffic state estimation. This study used a two-stage estimation approach, first estimating turning movements and then refining the estimates with observed data.
METHODOLOGY
The study developed a two-stage algorithm using a drone-based vehicle trajectory dataset from Orlando, Florida. It treated traffic signal optimization as a multi-player game where each signal phase represented a rational player. To estimate the payoff of each potential decision, upstream loop detector data was used to estimate vehicle arrivals. To evaluate the algorithm performance, the study used simulation software to model a four-legged intersection in downtown Toronto across varying CV market penetration levels (five to 100 percent).
FINDINGS
- The algorithm achieved up to a 37.7 percent reduction in average vehicle delay when using upstream density in its payoff function and a 34.1 percent reduction when using queue length.
- Standard deviation error for turning movement estimation decreased by up to 50 percent compared to other methods, with values improving from –0.14 to 0.70 at five percent market penetration and reaching 0.94 at 20 percent.
- Mean absolute percentage error for queues decreased by up to 45 percent and density estimates improved by up to 18.5 percent.
