Modeling influencing factors in a microscopic traffic simulator

Abstract

Microscopic traffic simulation is an important tool for traffic analysis and dynamic traffic management as it enables planners to evaluate traffic flow patterns, predict and evaluate the outcome of various response plans and assists in decision making. It is a vital tool for traffic management centers and can be helpful in developing contingency plans to enhance the safety and security of the transportation system. This thesis investigates the current state-of-the-practice in traffic microsimulation tools. A survey was developed and administered to developers. Results of the survey indicate critical gaps in including influencing external factors beyond the interaction of vehicles, such as incidents, work zones, or inclement weather, in traffic simulators. This thesis introduces a framework for incorporating such factors in existing models. The nature of the influencing factors limits disaggregate trajectory data collection generally needed to estimate driving behavior models. Therefore, an approach using aggregate calibration to refine and enhance existing driving behavior models is formulated. The aggregate calibration methodology is illustrated with a case study incorporating the effects of weather in driving behavior models using a freeway corridor in the Hampton Roads region of Virginia.

(cont.) MITSIMLab, a microscopic traffic simulation laboratory that was developed for evaluating the impacts of alternative traffic management system designs at the operational level, is used for evaluation. The presence of precipitation was found to be significant in reducing speeds in the case study and was incorporated into the driving behavior models with aggregate calibration. This methodology was found to improve the simulation results, by reducing bias and variability. Assessment of the approach is discussed and recommendations for improvement and further study are offered.