Human motion adapted for wheeled robots with arms and lifts
From Legs to Wheels: Embodiment-Aware Human Motion Retargeting for Mobile-Base Humanoids
Robotics
Summary
This paper addresses how to convert human walking motions into commands for robots that use wheels for movement instead of legs. The authors create a method that assigns parts of the human motion like walking and bending to different parts of the robot, like the base and arms. This helps the robot move in a way that follows the original human motion, even though it moves differently. They tested their method through simulations and running it on a real robot, showing it can successfully replicate human motion in wheeled humanoid robots.
What this means in practice
- •For robotics engineers: Translate human walking demonstrations into coordinated commands for wheeled humanoid robots with arms and lifts for better task execution.
- •For robotic system integrators: Implement motion retargeting to improve real-world deployment of mobile-base humanoid robots that must interpret human motion videos for control.
Authors
Jiyeon Koo, Eunseom Pyo, Jeonghee Seo, Taehwa Kim, Andrew Jaeyong Choi
Abstract
Human video offers a scalable source of robot demonstrations, yet most human-to-humanoid retargeting methods assume a legged robot with human-like kinematics. This assumption does not hold for mobile-base humanoids equipped with a wheeled base, vertical lift, and two arms. Human walking must be expressed through base motion, while torso bending may require coordinated lift and arm motion. We address this mismatch with a task-conditioned framework that assigns reconstructed human motion to base, lift, and arm responsibilities before robot-specific realization. The allocator preserves the human-derived path, stabilizes heading, separates turn and translation when needed, retimes commands to satisfy base limits, and repairs lift and arm trajectories. A deployment adapter then converts the reference to 50 Hz commands using stationary-base detection, deadband and slew-rate filtering, time-consistent playback scaling, and separate linear and angular gains. We evaluate the resulting references with human-derived task-space comparisons, policy-free simulation replay, and a qualitative execution on a physical robot.