Surrogate-Assisted High-Dimensional Optimization on Microscopic Traffic Simulators

Microscopic traffic simulation is able to capture many details of a traffic system, which makes it inherently interesting for simulation-based optimization. However, the considerable computational effort required for a single simulation run limits the use of standard heuristic optimization techniques and encourages the use of surrogate models to facilitate the search for an optimal solution. In this work, a grey-box surrogate model for microscopic traffic simulations is presented which allows the optimization of high-dimensional traffic optimization problems without relying on geographic or simulation-specific knowledge.