Global Optimization Approach for Deterministic, Stochastic and Robust Optimizations of Dynamic Integrated Network Design and Traffic Signal Setting Design Problem::Comparison and Evaluation Results

Abstract

The formulations of the user-optimal dynamic traffic assignment (UODTA)-based network design problem (NDP), signal setting design problem (SSD), and the integrated NDP and SSD problem (NDP-SSD) for the deterministic (DET), stochastic (SP) and robust (RO) optimizations are reviewed. The modifications of the simulated annealing (SA), genetic algorithm (GA) and reactive tabu search (RTS) for the nine problems (DET/SP/RO of NDP/SSD/NDP-SSD) are discussed. In the experiment, SA, GA and RTS are compared for the nine problems to identify best algorithms. The best algorithm for each problem is employed to find a best solution, which is then evaluated under stochastic condition. We find that the RO solutions yield the best robustness and lowest risk, whereas the DET solutions perform worst. The integrated robust approach is the most desirable, and the next best approach is the sequential robust approach. The deterministic approach can yield worse solutions than the do-nothing case.