Simulation-Based Optimization for Network Modeling With Heterogeneous Data

This chapter develops a simulation-based optimization (SBO) framework by integrating metamodels with mesoscopic simulation-based dynamic traffic assignment models for large-scale network modeling problems. The adopted SBO approach reconstructs the response surface by only a few evaluations of the objective function and is capable of handling simulation noises. The main benefit in computational timesavings can be achieved by using metamodels to construct response surfaces for predicting optimal solutions. This chapter provides a macroscopic understanding of urban traffic dynamics by using both a simulation-based dynamic traffic assignment model and heterogeneous traffic detection data. The simulation is validated by a representation of macroscopic fundamental diagrams (MFDs) using fixed traffic flow detections and probe travel time measurements. The SBO approach is demonstrated in a real-world large-scale transportation network that consists of arterials and freeways.