Decision-making based on reinforcement learning and model predictive control for highway on-ramp merging

Abstract

In recent years, autonomous driving cars have been attracting attention to reduce traffic accidents. Merging on the highway is one of the most challenging problems that need to be addressed for the realization of autonomous driving cars. The problem is difficult because an agent must decide where and how to merge under a complex changing environment. Reinforcement learning (RL) is one promising way for solving decision-making problems. However, it is hard to guarantee the safety of the controller obtained by RL. Therefore, we propose a combined method that decision making is done by RL and vehicle control by model predictive control to ensure safety. The performance of the proposed method is verified by simulation and shows a high success rate of merging.