RobotMover: Learning to Move Large Objects by Imitating the Dynamic Chain

2025-02-07

AI Summary

Robots face significant challenges in maneuvering large, heavy objects (e.g., furniture) within human environments—including whole-body motion coordination, complex underactuated dynamics modeling, and low-efficiency teleoperation. To address these, we propose RobotMover, a framework that introduces the “Dynamic Chain”: a spatially aware, velocity-sensitive topological structure linking the human root joint to the object’s root frame, enabling morphology-agnostic human-object dynamic imitation. This representation enables zero-shot cross-platform transfer without real-world fine-tuning. Our method integrates domain-randomized dynamic imitation learning in simulation, a dynamic consistency reward function, and a morphology redirection mechanism. Experiments demonstrate that RobotMover consistently outperforms both learning-based and teleoperation baselines across six quantitative metrics. It has been successfully deployed on physical hardware for real-world tasks, including garbage truck transportation and chair rearrangement.

Technology Category

• Intelligent Robots: Manipulation
• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Responsible Web: Machine-in-the-loop, human agency and autonomy
• User Modeling, Personalization and Recommendation: User modeling and simulation for interactive and conversational systems
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work

Abstract

Moving large objects, such as furniture, is a critical capability for robots operating in human environments. This task presents significant challenges due to two key factors: the need to synchronize whole-body movements to prevent collisions between the robot and the object, and the under-actuated dynamics arising from the substantial size and weight of the objects. These challenges also complicate performing these tasks via teleoperation. In this work, we introduce method, a generalizable learning framework that leverages human-object interaction demonstrations to enable robots to perform large object manipulation tasks. Central to our approach is the Dynamic Chain, a novel representation that abstracts human-object interactions so that they can be retargeted to robotic morphologies. The Dynamic Chain is a spatial descriptor connecting the human and object root position via a chain of nodes, which encode the position and velocity of different interaction keypoints. We train policies in simulation using Dynamic-Chain-based imitation rewards and domain randomization, enabling zero-shot transfer to real-world settings without fine-tuning. Our approach outperforms both learning-based methods and teleoperation baselines across six evaluation metrics when tested on three distinct object types, both in simulation and on physical hardware. Furthermore, we successfully apply the learned policies to real-world tasks, such as moving a trash cart and rearranging chairs.

Problem

• Synchronizing whole-body robot movements
• Handling under-actuated dynamics in objects
• Transferring human-object interaction to robots

Innovation

• Dynamic Chain representation
• Imitation learning framework
• Zero-shot transfer capability