A bi-level programming formulation and heuristic solution approach for traffic control optimization in networks with dynamic demand and stochastic route choice

This study develops a bi-level programming formulation and heuristic solution approach for traffic signal optimization in networks with time-dependent demand and stochastic route choice. In the bi-level programming model, the upper level problem represents the decision-making behavior (signal control) of transportation planner or system manager, while the user travel behavior is addressed in the lower level problem. The heuristic solution approach consists of a Genetic Algorithm (GA) and a Cell Transmission Simulation (CTS) based Incremental Logit Assignment (ILA) procedure, where GA is used to find the upper level signal control variables, while ILA is developed to find user optimal flow pattern in the lower level problem, and CTS is implemented to propagate traffic and collect real-time traffic information.; The essential part of the lower level problem is route choice model. This research proposes an algorithm framework which can accommodate various route choice models to investigate how the extended logit models change the equilibrium flow by using various treatments on overlapping paths and how much system performance can be improved under various congestion levels. The implementation of five logit models in two sample networks reveals that PCL produces the most significant changes in the equilibrium flow and the extended logit models lead to the improvement of system performance in terms of average travel time. The most widely used commercial simulation software, CORSIM, is used to validate the output of heuristic solution approach. The heuristic solution approach is applied in two signalized networks to search for the optimal signal control plan with the consideration of dynamic demand and stochastic route choice. To study the impact of different levels of ITS implementation in a transportation system, this research compares different information updating frequencies. In the numerical experiments, Elitist GA and Micro GA are compared and the impact of population size on the performance of GA is studied as well. It is noticed that Micro GA has more chance to achieve better results than Elitist GA using the same amount of fitness evaluation. The results also show that applying the optimal signal timing found by the heuristic solution approach can reduce the average travel time by 3∼8% in the test networks.