Residual Feedback Learning for Contact-Rich Manipulation Tasks with Uncertainty
Albert-Ludwigs-Universität Freiburg（阿尔伯特-路德维希-弗莱堡大学）；Karlsruhe Institute of Technology（卡尔斯鲁厄理工学院）；Bosch Center for Artificial Intelligence (BCAI)（博世人工智能中心（BCAI））
Alireza Ranjbar, Ngo Anh Vien, Hanna Ziesche, Joschka Boedecker, Gerhard Neumann
While classic control theory offers state of the art solutions in many problem scenarios, it is often desired to improve beyond the structure of such solutions and surpass their limitations. To this end, residual policy learning (RPL) offers a formulation to improve existing controllers with reinforcement learning (RL) by learning an additive "residual" to the output of a given controller. However, the applicability of such an approach highly depends on the structure of the controller. Often, internal feedback signals of the controller limit an RL algorithm to adequately change the policy and, hence, learn the task. We propose a new formulation that addresses these limitations by also modifying the feedback signals to the controller with an RL policy and show superior performance of our approach on a contact-rich peg-insertion task under position and orientation uncertainty. In addition, we use a recent Cartesian impedance control architecture as the control framework which can be available to us as a black-box while assuming no knowledge about its input/output structure, and show the difficulties of standard RPL. Furthermore, we introduce an adaptive curriculum for the given task to gradually increase the task difficulty in terms of position and orientation uncertainty. A video showing the results can be found at https://youtu.be/SAZm_Krze7U .
NoContactNoWorries: Estimating Contact through Vision and Proprioception for In-Hand Dexterous Manipulation
Robotics Research Center (RRC), International Institute of Information Technology (IIIT), Hyderabad, India（印度海得拉巴国际信息技术研究所机器人研究中心）；Department of Computer Science, The University of Manchester（曼彻斯特大学计算机科学系）
Maglev-Pentabot: Magnetic Levitation System for Non-Contact Manipulation using Deep Reinforcement Learning
Department of Electrical and Computer Engineering, University of Wisconsin-Madison（电气与计算机工程系，威斯康星大学麦迪逊分校）；Flexcompute；Department of Electrical Engineering, Stanford University（电气工程系，斯坦福大学）
Residual Learning from Demonstration: Adapting DMPs for Contact-rich Manipulation
University of Edinburgh（爱丁堡大学）；Aalto University（阿尔托大学）；Monash University（莫纳什大学）；Intrinsic（Intrinsic公司）；Facebook AI Research（脸书人工智能研究院）
涉及接触和摩擦的操作技能是许多机器人任务所固有的。使用类的电机原语钉孔一样插入，我们研究机器人如何学习这些技能。动态运动原语（Dynamic Movement Primitives，缩写为DMOR）是一种通过行为克隆（Behavior Cloning，缩写为BC）提取此类策略的流行方法，但在插入上下文中可能会遇到困难。策略自适应策略，如剩余学习，可以帮助提高策略的整体性能，在接触丰富的操纵。然而，目前尚不清楚如何最好地使用DMP。因此，我们考虑了几种可能的方法来适应一个学习的公式，并提出了“残余示教学习”（rLfD），一个框架，结合DMPS与强化学习（RL）学习残余校正策略。我们的评估表明，直接在任务空间中应用残差学习并在机器人的完整姿势上操作可以显着提高DMPs的整体性能。我们发现，rLfD提供了一个温和的关节解决方案，提高了任务的成功和推广的DMPS \rb{and enables transfer to different geometries and frictions through few-shot task adaptation}。根据一系列任务对拟议框架进行评价。一个模拟机器人和一个物理机器人必须成功地将钉子、齿轮和插头插入各自的插座。本文附带的其他材料和视频可在https://sites.google.com/view/rlfd/上找到。
Approximately Optimal Global Planning for Contact-Rich SE(2) Manipulation on a Graph of Reachable Sets
Robotics Institute, Carnegie Mellon University（卡内基梅隆大学机器人学院）；Robotics and AI Institute（机器人与人工智能研究所）；CSAIL, Massachusetts Institute of Technology（麻省理工学院计算机科学与人工智能实验室）
Push, Press, Slide: Mode-Aware Planar Contact Manipulation via Reduced-Order Models
Dept. of Electrical and Electronics Engineering, Middle East Technical University（电子工程系，中东技术大学）；Dept. of Computer Engineering, Bilkent University（计算机工程系，比尔肯特大学）
Aerial Manipulation: Contact, Medium Coupling, and the Geometry of Readiness
Robotics and Mechatronics lab, Faculty of Electrical Engineering, Mathematics & Computer Science, University of Twente（特温特大学电气工程、数学与计算机科学学院机器人与机电一体化实验室）；Department of Computer, Control and Management Engineering, Sapienza University of Rome（罗马第一大学计算机、控制与管理工程系）
ManiFoundation Model for General-Purpose Robotic Manipulation of Contact Synthesis with Arbitrary Objects and Robots
School of Electronics Engineering and Computer Science, Peking University（北京大学电子工程与计算机科学学院）；National University of Singapore（新加坡国立大学）；Shanghai Jiao Tong University（上海交通大学）
Contact SLAM: An Active Tactile Exploration Policy Based on Physical Reasoning Utilized in Robotic Fine Blind Manipulation Tasks
Research Center for Intelligent Robotics（智能机器人研究中心）；National Key Laboratory of Aerospace Flight Dynamics（航空飞行动力学国家重点实验室）；Shaanxi Province Innovation Team of Intelligent Robotic Technology（陕西省智能机器人技术创新团队）；School of Astronautics, Northwestern Polytechnical University（航空学院，西北工业大学）
Should We Learn Contact-Rich Manipulation Policies from Sampling-Based Planners?
New York University（纽约大学）；Boston Dynamics AI Institute（波士顿动力AI研究院）；Georgia Institute of Technology（佐治亚理工学院）；Cornell University（康奈尔大学）；Artificial and Natural Intelligence Toulouse Institute（图卢兹人工智能与自然智能研究所）
