A Model for Real-Time Traffic Coordination Using Simulation Based Optimization

An approximate traffic model, APRES-NET, is proposed for a simulation-based optimization approach for real-time network coordination of traffic flows. The objective of APRES-NET is twofold: to provide network-wide prediction of traffic movements and to evaluate, approximately but quickly, the performance of any signal timing strategy. To make a prediction, it uses the observed network vehicular traffic and signal data for a period of time, a future signal plan, and statistical information on vehicle movement, to predict the traffic for some future time period. The APRES-NET simulation consists of propagating into the future the movement of the vehicles detected in the network during the “last detection period”. To quickly, and approximately, simulate the vehicle movement through the whole network, a simplified queuing mechanism is implemented at the intersections. At an intersection, a turn is assigned to each vehicle based on turning probabilities. Performance measures are calculated continuously as the simulation run is being conducted. APRES-NET is being used for real-time coordination of signal phase timings by making it behave as a “function evaluator” in an iterative optimization scheme. The paper discusses the basic structure and the algorithms used by APRES-NET.