A novel control strategy of automatic parallel parking system based on Q-learning

Abstract

Parallel parking is the most common form of parking, but it is difficult to park vehicles safely in complex traffic scenes. The paper proposed an automatic parallel parking control strategy based on Q-learning. The convergent Q-table was obtained through offline training to ensure that the automatic parking control system could obtain the optimal parking action during the automatic parking process. Based on the parking kinematics model, the virtual parking scene and the parallel parking control strategy based on fuzzy control were established. The comparative analysis of the two control strategies showed that the parallel automatic parking control strategy based on a Q-learning algorithm could deal with more complex traffic scenes. Finally, the automatic parallel parking hardware-in-the-loop test system was built based on a purely electric drive platform, and the effectiveness of the automatic parking control strategy was further verified through the hardware-in-the-loop test.