Simulation Based Methodology for Assessing Forced Merging Strategies for Autonomous Vehicles

A simulation environment for assessing the forced merging strategies for autonomous vehicles (AV) is described. Such forced merging strategies need to explicitly account for vehicle interactions and require extensive simulation tests before deployment. The proposed environment facilitates such simulation tests to better assess, validate, and improve the forced merging strategies. The simulation environment consists of several control strategies for the merging vehicle, different highway vehicle models that can interact with the merging vehicle, and a set of metrics to evaluate the performance of the merging strategies. As an example of the evaluation, a Leader Follower Game Controller (LFGC) [1] is implemented and tested in this environment. Based on the analysis of the simulation tests, we propose an Enhanced LFGC (e-LFGC). It is shown that our methodology expedites the development of merging controllers so that corner cases are revealed and control requirements can be improved.