Robotics

Universal Manipulation Exoskeleton: Learning Compliant Whole-body Policies with Real-time Torque Feedback

James (Jingxi) Xu, Mark Cutkosky

Featured June 24, 2026

AI-generated analysis — This is SciGrove's AI interpretation of the paper, not peer-reviewed content. Always refer to the original paper.

Simply

A special arm-worn device lets people 'feel' what a robot arm is touching, helping them teach robots to safely handle tricky tasks that require gentle pushing and pulling, even in tight spaces.

In depth

The paper introduces UME, an exoskeleton that provides real-time haptic torque feedback and captures whole-arm configurations. This system enables human operators to intuitively teleoperate diverse robots for complex, contact-rich tasks. Crucially, it facilitates the collection of rich force and torque data, allowing the learning of active compliant policies that adapt to physical interactions, a capability often lacking in prior methods.

Key Takeaways

1
UME provides real-time haptic torque feedback, allowing operators to 'feel' robot interactions and enabling the collection of crucial torque data for learning compliant behaviors.
2
A universal retargeting algorithm allows UME to intuitively control diverse robot arms with varying degrees of freedom, enhancing system stability by decomposing the kinematic chain into sub-manipulators.
3
The system's portability and embedded IMU facilitate teleoperation for mobile manipulation, enabling the learning of robust, long-horizon, whole-body compliant policies for complex tasks.

Conceptual Flow

HIGH LEVEL

Methodology: Teaching Robots by Feeling

A person wears a special arm device that lets them control a robot and feel what the robot touches, helping the robot learn how to move carefully.

Human Arm Moves

Exoskeleton Senses

Robot Arm Moves

Robot Feels Forces

Robot Learns Safe Actions

Results: Robots Learn to Succeed

Without feeling forces, robots often fail tricky tasks, but with the new device, they learn to succeed by feeling and reacting.

Robot without Force Sense

Fails Tricky Tasks

Robot with Force Sense

Succeeds Tricky Tasks

This breakdown was generated by SciGrove. Get the same analysis — intuition, storyboard, peer review, a runnable prototype and a glossary — on any paper you upload or paste a DOI for.