Title:Discretionary Lane-Change Decision and Control via Parameterized Soft Actor-Critic for Hybrid Action Space

Abstract:This study focuses on a crucial task in the field of autonomous driving, autonomous lane change. Autonomous lane change plays a pivotal role in improving traffic flow, alleviating driver burden, and reducing the risk of traffic accidents. However, due to the complexity and uncertainty of lane-change scenarios, the functionality of autonomous lane change still faces challenges. In this research, we conduct autonomous lane-change simulations using both Deep Reinforcement Learning (DRL) and Model Predictive Control (MPC). Specifically, we propose the Parameterized Soft Actor-Critic (PASAC) algorithm to train a DRL-based lane-change strategy to output both discrete lane-change decision and continuous longitudinal vehicle acceleration. We also use MPC for lane selection based on predictive car-following costs for different lanes. For the first time, we compare the performance of DRL and MPC in the context of lane-change decision. Simulation results indicate that, under the same reward/cost functions and traffic flow, both MPC and PASAC achieve a collision rate of 0\%. PASAC demonstrates comparable performance to MPC in terms of episodic rewards/costs and average vehicle speeds.