首页|New Robotics and Automation Study Findings Have Been Reported by Researchers at Xi'an Jiaotong University (Task-driven Autonomous Driving: Balanced Strategies I ntegrating Curriculum Reinforcement Learning and Residual Policy)
New Robotics and Automation Study Findings Have Been Reported by Researchers at Xi'an Jiaotong University (Task-driven Autonomous Driving: Balanced Strategies I ntegrating Curriculum Reinforcement Learning and Residual Policy)
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
By a News Reporter-Staff News Editor at Robotics & Machine Learning DailyNews Daily News-Investigators discuss new findings in Robotics - Robotics and Automation. Accordingto news originating from Xi'an, Pe ople's Republic of China, by NewsRx correspondents, research stated,"Achieving fully autonomous driving in urban traffic scenarios is a significant challenge t hat necessitatesbalancing safety, efficiency, and compliance with traffic regul ations. In this letter, we introduce a novelCurriculum Residual Hierarchical Re inforcement Learning (CR-HRL) framework."Financial supporters for this research include National Key R&D Pro gram of China, National NaturalScience Foundation of China (NSFC).Our news journalists obtained a quote from the research from Xi'an Jiaotong Univ ersity, "It integrates arule-based planning model as a guiding mechanism, while a deep reinforcement learning algorithm generatessupplementary residual strate gies. This combination enables the RL agent to perform safe and efficientoverta king in complex traffic scenarios. Furthermore, we implement a detailed three-st age curriculumlearning strategy that enhances the training process. By progress ively increasing task complexity, thecurriculum strategy effectively guides the exploration of autonomous vehicles and improves the reusabilityof sub-strategi es. The effectiveness of the CR-HRL framework is confirmed through ablation expe riments.Comparative experiments further highlight the superior efficiency and d ecision-making capabilities of ourframework over traditional rule-based and RL baseline methods."
Xi'anPeople's Republic of ChinaAsiaRobotics and AutomationRoboticsEmerging TechnologiesMachine LearningRei nforcement LearningXi'an Jiaotong University
NETL
NSTL
