Calibrating simulation models for Advanced Traveler Information Systems/Advanced Traffic Management Systems applications

This paper is an attempt to address the issue of data flow interdependency and error propagation in Advanced Traveler Information Systems/Advanced Traffic Management Systems supporting simulation-based models. A complex dynamic simulation model constitutes a number of processes, variables, or parameters with varying impact on the model's state estimation accuracy. The overall model error might be attributed to the propagating internal system errors. This paper presents a framework for calibrating simulation models developed to support the dynamic traffic assignment systems. The calibration framework employs off-line analyses on the model processes, input, and system data, and the results are utilized to identify the model processes causing significant errors and to quantify the propagation of a specific variable error. The framework could be utilized to rank the model processes and their variables so that resources are allocated to calibrating significant error-causing entities. The framework is tested on a microscopic simulation model, MTSSIMA, and the results are discussed.