In this paper, a method for autonomous segmentation of demonstrated robot movements is proposed. Position data is clustered into Gaussian mixture models (GMMs), and an initial set of segments is identified from the Gaussian basis functions. A Kalman filter is used to detect sudden changes in the contact force/torque measurements, and this is used to update and verify the initial segmentation points. The segmentation method is verified experimentally on an industrial robot.
Hierarchical Planning of Dynamic Movements without Scheduled Contact Sequences
Department of Advanced Robotics, Istituto Italiano di Tecnologia(意大利技术研究院先进机器人部);Robot Learning and Interaction Group, Idiap Research Institute(伊迪普研究所机器人学习与交互小组)
Deep Reinforcement Learning for Contact-Rich Skills Using Compliant Movement Primitives
Faculty of Mechanical Engineering(机械工程学院)
近年来,工业机器人已被安装在各个行业,以处理先进制造和高精度任务。然而,与人类操作员相比,工业机器人的进一步集成受到其有限的灵活性,适应性和决策技能的阻碍。装配任务对机器人来说尤其具有挑战性,因为它们具有丰富的接触性,对即使很小的不确定性也很敏感。虽然强化学习(RL)提供了一个很有前途的框架,从头开始学习接触丰富的控制策略,其适用性高维连续状态动作空间仍然相当有限,由于高脆性和样本的复杂性。为了解决这些问题,我们提出了不同的修剪方法,促进收敛和推广。特别是,我们将任务分为自由和接触丰富的子任务,在笛卡尔空间而不是关节空间中执行控制,并参数化控制策略。这些修剪方法自然地在动态运动基元(dynamic movement primitives,简称HMM)的框架内实现。为了处理接触丰富的任务,我们通过引入一个耦合项来扩展RISK框架,该耦合项的作用类似于人类手腕,并在与环境接触时提供主动顺应性。我们证明了所提出的方法可以学习对空间,大小,形状和密切相关的场景不变的插入技能,同时处理大的不确定性。最后,我们证明了学习的政策可以很容易地从模拟转移到现实世界中,并实现类似的性能UR5e机器人。
TacUMI: A Multi-Modal Universal Manipulation Interface for Contact-Rich Tasks
School of Computation, Information and Technology, Technical University of Munich(技术大学慕尼黑计算、信息与技术学院);Agile Robots SE(敏捷机器人公司);State Key Laboratory for Novel Software Technology and the School of Science and Technology, Nanjing University (Suzhou Campus)(南京大学软件新技术国家重点实验室及科学与技术学院(苏州校区));Shanghai University(上海大学)
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)