基于封闭场景自动驾驶的DDPG决策规划方法

打开文本图片集

A DDPG-Based Planning Method for Autonomous Driving in Confined Environments

Jiang LiweilWang Manjiang1Qiu Qian1Xiao WenchaolZhou BolZhu Hailong² 1.Wuhan Bus Manufacturing Co.，Ltd.，Technology Center，Wuhan 430200，China 2.Military Representative Bureau of Chemical Defense，Army Equipment Department，Yichang 443000，China

Abstract：With the advancementof autonomousdriving technology，decision-makingand planing inunstructured road scnarios havebecomearesearchfocus.Traditionalrule-basedmethodsexhibit limited generalizationcapabilityincomplexanddynamicenviron ments，struglingtobalanceeficiencyandeconomy.Besedonthis，the paperproposesadecision-makingand planning methodbased onDep DeterministicPolicyGradient（DDPG）.Byreasonablydesigning the stateandactionspaceandconstructingareward function that ntegrateseficiencyconomy，afety，andcomfort，simulationexperimentsreconductedinustructuredroadscenariosiningareas.Theresultsdemonstratethattheproposed methodachievesexcellentperformanceacross multipleaspects，signicantlyenhancing evehicle'sdecision-makingandplanningcapabilies，andprovidinganovelapproachfortheappicationofautonomousdriing in complex road environments.

Key Words：Decision planning;DDPQ;Unstructured roads scenarios;Mining scene

1前言

自动驾驶系统的决策规划模块负责解析环境信息并生成安全高效的行驶指令，其性能直接影响车辆在复杂交通环境中的可靠性[1-2]。（剩余7076字）