A Stochastic Simulation-Based Optimization Method for Calibrating VISSIM Simulator under Uncertainties

Simulation plays a fundamental role in evaluating traffic operations and transportation planning strategies. A reliable simulator can provide effective analysis of a given traffic network if the simulation model parameters are accurately calibrated to local conditions. To solve the calibration problem under changes in traffic demand, this paper constructs a simulation-based optimization (SO) platform and proposes a stochastic simulation-based optimization (SSO) approach. The regressing Kriging model is adopted to filter out noise generated from random traffic demand. Through algorithm comparison for a benchmark stochastic optimization problem, it is validated that SSO performs better than GA with the same computation cost. In the field experiment, an urban artery network with four signalized intersections in Nanchang, China is modeled by VISSIM. Results indicate that over the morning peak the trend of simulation volume from the calibrated model resembles the actual traffic volume. The proposed SSO approach can address black-box problems under uncertainties.