电子科技2026，Vol.39Issue(1)：57-63,7.DOI:10.16180/j.cnki.issn1007-7820.2026.01.008
基于IL-TD3的轮式机器人无地图导航研究
Research on Mapless Navigation of Wheeled Robot Based on IL-TD3
摘要
Abstract
In view of the dependence of the wheeled robot navigation method on high-precision maps and its a-daptability issues in dynamic and complex scenarios,this study proposes a mapless navigation method based on IL-TD3(Imitation Learning Enhanced Twin Delayed Deep Deterministic Policy Gradient).The navigation task is modeled as a POMDP(Partially Observable Markov Decision Process),and a LSTM(Long Short-Term Memory)network is combined to process historical information for improving the environmental state modeling.The robot quickly acquires limited mapless navigation capabilities through imitation learning,and continuously explores and trains the TD3 deep reinforcement learning network to enhance its navigation skills.The simulation experiment results show that the naviga-tion trajectory of IL-TD3 in unknown dynamic environments is stable,continuous,and safe,demonstrating good navi-gation performance.The Sim2Real(Simulation to Reality)test results indicate that the unadjusted IL-TD3 model per-forms well in real-world navigation tasks,which proves the robustness and generalization ability of the proposed model.关键词
机器人/无地图导航/POMDP/LSTM/模仿学习/深度强化学习/未知动态环境/Sim2RealKey words
robot/mapless navigation/POMDP/LSTM/imitation learning/deep reinforcement learning/unknown dyna-mic environments/Sim2Real分类
信息技术与安全科学引用本文复制引用
秦源赛,牟海明,刘元基,李清都..基于IL-TD3的轮式机器人无地图导航研究[J].电子科技,2026,39(1):57-63,7.基金项目
东方学者计划(TP2019064)Oriental Scholars Program(TP2019064) （TP2019064）
