DRBO - A Simulator Calibration Framework Based on Day-to-Day Dynamic Routing and Bayesian Optimization

Traffic simulation, a tool for recreating real-life traffic scenarios, acts as an important platform in transportation research. Considering the growing complexity of urban mobility, various large scale simulators are designed and used for research and applications. This paper proposes DRBO, a calibration framework for large scale traffic simulators. This framework combines the travel behavior adjustment with Bayesian Optimization, better exploring the structure of the simulator as well as improving its performance. By the calibration procedure, we decrease the gap between the simulator output and the real data, making the system much more reliable. Compared to the prior art, our framework is efficient for capturing multiple perspectives of the simulator, using Bayesian optimization and rerouting techniques simultaneously, achieving a KL distance value of 0.053 between simulated speed and observed speed, and we further tested our simulator on SFCTA demand to further validate the speed distribution from our simulation and observed data.