Humanoids learn to vary object transport distance from one motion clip
Learning Distance-Conditioned Object Transport for Humanoid Loco-Manipulation from a Single Motion Clip
Robotics
Summary
Humanoid robots can move objects by copying human motions, but usually only imitate the final position shown in a single example. The authors found that intermediate steps in the movement aren’t treated as goals, which limits flexibility. They created a new method that reuses parts of the motion to teach the robot how to reach different distances. This approach helps the robot adapt its movements to new transport distances and improves performance both in simulations and on real robots.
What this means in practice
- •For robotics engineers: Develop humanoid robots that can flexibly transport objects to varying distances by learning from a single example motion.
- •For automation system integrators: Integrate adaptable humanoid object manipulation into complex tasks requiring precise control over transport distances.
Authors
Yuhyeon Hwang, Daniel Sungho Jung, YongHyeok Seo, Mingi Jung, Chang Nho Cho, Jung-Hoon Hwang, Dongin Shin
Abstract
Motion tracking can reproduce humanoid loco-manipulation from a single retargeted motion clip, but a policy trained on a fixed reference primarily reproduces its demonstrated transport outcome. Although the source trajectory visits intermediate object displacements, transport termination is demonstrated only at its endpoint. We identify this mismatch as the termination-versus-passage gap: intermediate displacements are observed as passage states rather than termination-complete outcomes. We introduce Distance-Conditioned Reference Recomposition (DCRR), which relocates the demonstrated termination segment to intermediate transport states. A frozen tracking teacher replays the recomposed references under closed-loop dynamics, and the retained trajectories are relabeled by their achieved object placements and distilled into a reference-free policy. This procedure constructs distance-conditioned supervision from the interaction behavior encoded in the source motion. Across Carry, Kick-Push, Crouch-Push, and Drag, DCRR-BC produces command-dependent transport with an overall normalized distance mean absolute error (MAE) of 0.15, compared with 0.28 for source-only behavior cloning. RL fine-tuning further improves the command response and execution robustness in the training simulator and under sim-to-sim transfer. Finally, hardware experiments demonstrate transport-distance modulation across all four interaction modes.