🤖🍨 Sundae Robotics 01: Robotics, Dexterity, Cross-Embodiment & World Models — Universal Manipulation Exoskeleton (UME)
🤖🍨 Grab a sundae and join Sundae Robotics, a private, invite-only Sunday series bringing together robotics researchers, founders, and builders working at the frontier of physical intelligence.
Sundae Robotics 01
Robotics, Dexterity, Cross-Embodiment & World Models
Featured Talk: Universal Manipulation Exoskeleton (UME)
Keynote: James (Jingxi) Xu
Senior Research Scientist, Ant Group · Stanford BDML (Mark Cutkosky) · Former Boston Dynamics AI Institute · PhD, Columbia (Matei Ciocarlie & Shuran Song)
For robots to work safely in household environments, they need to be compliant and react to torque and force feedback during contact. However, the majority of existing data collection pipelines still lack the ability to capture force and torque data for learning active compliant policies.
In this talk, James will present Universal Manipulation Exoskeleton (UME), a new groundbreaking 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 even unsheathe kinematically constrained objects while blindfolded. UME is low-cost, lightweight, and portable. Equipped with an embedded IMU, it enables teleoperation for mobile manipulation. With its proposed universal retargeting algorithm, UME can teleoperate a range of robots, including the 7DoF OpenArm, 7DoF Franka, and 6DoF X-ARM. We demonstrate that this combination of capabilities enables learning bimanual, whole-body, and active compliant policies that operate effectively in highly constrained spaces. The learned robust autonomous policies achieve high success rates across a variety of tasks, including long-horizon mobile manipulation, force-mediated box flipping, visually occluded box pushing, and space-constrained tabletop manipulation. James will also bring the UME exoskeleton for attendees to see up close following the talk.
James's research spans robot learning, tactile sensing, dexterous manipulation, and rehabilitation robotics, with publications at ICRA, IROS, CoRL, RSS, RA-L, AuRo, and CASE. He is an RSS Pioneer 2026 and has given invited talks at Stanford Vision Lab (hosted by Fei-Fei Li), Ai2, Meta, Amazon FAR, Amazon Robotics, and the RAI Institute.
Pre-Reading
• Universal Manipulation Exoskeleton: Learning Compliant Whole-Body Policies with Real-Time Torque Feedback (2026)
Topics
• Haptic torque feedback and compliant manipulation
• Whole-arm teleoperation and cross-embodiment retargeting
• Contact-rich robot learning
• Robot foundation models with force, torque, proprioception, and tactile sensing
Open Discussion + Q&A
Is force/torque the missing modality for robot foundation models?
Should robot policies predict actions, forces, or impedance?
What is the best interface for collecting manipulation data at scale?