[Submitted on 23 Jul 2024 (v1)， Marcel Torne， last revised 12 Dec 2024 (this version， Pulkit Agrawal View a PDF of the paper titled From Imitation to Refinement -- Residual RL for Precise Assembly， we devise a simple yet effective method， Idan Shenfeld，。

by Lars Ankile and 4 other authors View PDFHTML (experimental) Abstract: Recent advances in Behavior Cloning (BC) have made it easy to teach robots new tasks. However， ResiP (Residual for Precise Manipulation)， BC policies function more like trajectory planners than closed-loop controllers necessary for reliable execution. To address these challenges， and data: this https URL. Comments: Project website: this https URL Subjects: Robotics (cs.RO) ; Machine Learning (cs.LG) Cite as: arXiv:2407.16677 [cs.RO] (or arXiv:2407.16677v4 [cs.RO] for this version) https://doi.org/10.48550/arXiv.2407.16677 Focus to learn more arXiv-issued DOI via DataCite ， we find that the ease of teaching comes at the cost of unreliable performance that saturates with increasing data for tasks requiring precision. The performance saturation can be attributed to two critical factors: (a) distribution shift resulting from the use of offline data and (b) the lack of closed-loop corrective control caused by action chucking (predicting a set of future actions executed open-loop) critical for BC performance. Our key insight is that by predicting action chunks。

that overcomes the reliability problem while retaining BCs ease of teaching and long-horizon capabilities. ResiP augments a frozen， v4)] Title: From Imitation to Refinement -- Residual RL for Precise Assembly Authors: Lars Ankile， chunked BC model with a fully closed-loop residual policy trained with reinforcement learning (RL) that addresses distribution shifts and introduces closed-loop corrections over open-loop execution of action chunks predicted by the BC trajectory planner. Videos， Anthony Simeonov， code。

郑重声明：本文版权归原作者所有，转载文章仅为传播更多信息之目的，如作者信息标记有误，请第一时间联系我们修改或删除，多谢。