Abstract

To re-establish regular operations in a tram traffic network after a large disturbance, e.g. resulting from vehicle breakdown or station closure, the viability of several rescheduling and rerouting strategies has to be evaluated prior to their implementation. Here, a multi-modal traffic simulation system can help to enhance the decision quality. Such a system obviously faces tight time constraints, so simulation data has to be acquired fast. In this paper we propose a method for the parallel execution of discrete traffic simulation models, which would accelerate data generation in comparison to a sequential model. To assess this method's dynamic behavior in real-world applications, some experiments conducted on a software system modeling schedule based tram traffic are presented.
After giving an introduction to the scope and aim, we show some background on the parallelization of discrete simulation models. The main part of the paper begins with the proposal of a method to parallelize the execution of simulation models with problem specific properties. Some estimations of the method's efficiency are shared, followed by several experiments to highlight its dynamic behavior in real-world applications. The paper ends with a short summary and some thoughts on further research.