首页|South China University of Technology Reports Findings in Robotics (Extended resi dual learning with one-shot imitation learning for robotic assembly in semi-stru ctured environment)
South China University of Technology Reports Findings in Robotics (Extended resi dual learning with one-shot imitation learning for robotic assembly in semi-stru ctured environment)
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 Daily News Daily News-New research on Robotics is the subjec t of a report. According to news originating from Guangzhou, People's Republic o f China, by NewsRx correspondents, research stated, "Robotic assembly tasks requ ire precise manipulation and coordination, often necessitating advanced learning techniques to achieve efficient and effective performance. While residual reinf orcement learning with a base policy has shown promise in this domain, existing base policy approaches often rely on hand-designed full-state features and polic ies or extensive demonstrations, limiting their applicability in semi-structured environments." Our news journalists obtained a quote from the research from the South China Uni versity of Technology, "In this study, we propose an innovative Object-Embodimen t-Centric Imitation and Residual Reinforcement Learning (OEC-IRRL) approach that leverages an object-embodiment-centric (OEC) task representation to integrate v ision models with imitation and residual learning. By utilizing a single demonst ration and minimizing interactions with the environment, our method aims to enha nce learning efficiency and effectiveness. The proposed method involves three ke y steps: creating an object-embodiment-centric task representation, employing im itation learning for a base policy using via-point movement primitives for gener alization to different settings, and utilizing residual RL for uncertainty-aware policy refinement during the assembly phase. Through a series of comprehensive experiments, we investigate the impact of the OEC task representation on base an d residual policy learning and demonstrate the effectiveness of the method in se mi-structured environments. Our results indicate that the approach, requiring on ly a single demonstration and less than 1.2 h of interaction, improves success r ates by 46% and reduces assembly time by 25%."
GuangzhouPeople's Republic of ChinaA siaEmerging TechnologiesMachine LearningReinforcement LearningRoboticsRobots
NETL
NSTL
