In household environments, robots need to be compliant and responsive to torque and force feedback during contact to ensure safety. However, most existing data collection pipelines lack the ability to capture torque and force data for learning active compliant policies. We present the Universal Manipulation Exoskeleton (UME), an upper-limb exoskeleton that provides real-time haptic torque feedback while recording whole-arm configurations and joint torque signals for teleoperation. With transparent torque feedback, human operators can manipulate kinematically constrained objects even when blindfolded. UME is low-cost, lightweight, and portable, equipped with an embedded IMU, enabling teleoperation for mobile manipulation. Our proposed universal retargeting algorithm allows UME to teleoperate a range of robots, including the 7DoF OpenArm, 7DoF Franka, and 6DoF X-ARM. We demonstrate that this combination of capabilities enables the learning of bimanual, whole-body, and active compliant policies that operate effectively in highly constrained spaces. The learned robust autonomous policies achieve high success rates across various tasks, including long-horizon mobile manipulation, force-mediated box flipping, visually occluded box pushing, and space-constrained tabletop manipulation. Videos, code, and additional information can be found at UME official website.
Blogger's Review: UME significantly enhances robotic manipulation capabilities in complex environments through real-time torque feedback and flexible retargeting algorithms. This technology not only improves safety and adaptability but also paves the way for future smart home applications.