Partial Motion Imitation for Learning Cart Pushing with Legged Manipulators

arXiv cs.RO / 3/30/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper addresses the challenge of learning stable locomotion while executing precise manipulation for legged robots in real-world mobile manipulation tasks like pushing carts.
It introduces a partial imitation learning method that transfers locomotion style by first training a robust locomotion policy with extensive domain and terrain randomization.
For loco-manipulation, it learns a separate policy by imitating only lower-body motions using a partial adversarial motion prior, rather than imitating full-body behavior.
Experiments show the resulting policy can push a cart along diverse trajectories in IsaacLab and can transfer effectively to MuJoCo.
Compared with multiple baselines, the approach improves both stability and accuracy of loco-manipulation behaviors.

Abstract

Loco-manipulation is a key capability for legged robots to perform practical mobile manipulation tasks, such as transporting and pushing objects, in real-world environments. However, learning robust loco-manipulation skills remains challenging due to the difficulty of maintaining stable locomotion while simultaneously performing precise manipulation behaviors. This work proposes a partial imitation learning approach that transfers the locomotion style learned from a locomotion task to cart loco-manipulation. A robust locomotion policy is first trained with extensive domain and terrain randomization, and a loco-manipulation policy is then learned by imitating only lower-body motions using a partial adversarial motion prior. We conduct experiments demonstrating that the learned policy successfully pushes a cart along diverse trajectories in IsaacLab and transfers effectively to MuJoCo. We also compare our method to several baselines and show that the proposed approach achieves more stable and accurate loco-manipulation behaviors.