Open AccessAn Automatic Parking Model Based on Deep Reinforcement Learning
Author(s) -
Li Junzuo,
Qiang Liu
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1883/1/012111
Subject(s) - heading (navigation) , generalization , reinforcement learning , process (computing) , computer science , displacement (psychology) , kinematics , point (geometry) , function (biology) , focus (optics) , action (physics) , artificial intelligence , reinforcement , line (geometry) , simulation , engineering , mathematics , structural engineering , psychology , mathematical analysis , physics , geometry , optics , classical mechanics , quantum mechanics , evolutionary biology , psychotherapist , biology , aerospace engineering , operating system
When parking a car, it is crucial to ensure the car constantly approaches the parking point, gets an excellent heading angle, and avoids significant losses caused by line pressure. An automatic parking model based on deep reinforcement learning is proposed. A parking kinematics model is built to calculate the different states of its movement. Steering angle and displacement are used to achieve interaction as actions; A comprehensive reward function is designed to consider the focus of action and safety in different parking stages. Through training, the car’s automatic parking is realized, and a comprehensive analysis of the various stages and situations in the parking process is given. Besides, it is showed by a further generalization experiment that the model has good generalization.