Validation of High-Fidelity Traffic Simulation Models

A multistage validation framework that accounts for the realistic nature of traffic simulation output data is proposed. The framework consists of conceptual validation and operational validation. The operational validation comprises a qualitative approach, which involves static and animated Turing tests, and a quantitative approach, which involves three levels of statistical tests. Particularly in the third-level statistical test, the autocorrelation and nonstationary nature of traffic simulation output data is emphasized, its implications on validation methods are explored, and a univariate nonseasonal autoregressive-integrated-moving-average (ARIMA) modeling approach is proposed. Finally, numerical results for an actual freeway network are presented. The validation results illustrate that the proposed multistage validation procedure can account for the complexity of the validation task and its conclusions.