KAI: A Kinematic-Aware Interface for Data-Efficient Articulated Object Manipulation

2026-07-27

AI Summary

This work addresses the challenge of inefficiently learning the kinematic structures of articulated objects from robot demonstrations alone. To overcome this limitation, the authors propose KAI, a structured intermediate representation that embeds geometric and kinematic priors into policy learning to align with the object’s intrinsic motion structure. KAI introduces an action-agnostic, kinematics-aware interface that enables joint training with human demonstration videos and exhibits strong generalization capabilities, enabling stable manipulation even under unseen backgrounds and visual distractions. Experimental results demonstrate that, using only half the amount of demonstration data, KAI achieves an average success rate of 82.9% across six simulated tasks, significantly outperforming baseline methods. Furthermore, when trained jointly with real-world human–robot interaction videos, KAI attains an average success rate exceeding 70% in cluttered environments.

Abstract

Articulated object manipulation requires an understanding of kinematic structure that is difficult and costly to learn from robot demonstrations alone. We introduce the Kinematic-Aware Articulation Interface (KAI), a structured intermediate representation that captures the kinematic structure of articulated objects. By embedding interpretable geometric and kinematic priors into policy learning, KAI provides a strong inductive bias aligned with the underlying structure of articulated motion. This design effectively improves sample efficiency, with gains particularly pronounced in low-data regimes: across six simulation tasks, our method achieves an average success rate of 82.9%, matching or surpassing baseline performance while using only half the demonstration data. Our method also exhibits robust generalization to unseen backgrounds and visual distractors, transferring from a single clean training environment to cluttered real-world scenes. KAI's action-agnostic design further enables co-training with human interaction videos to enhance real-world robustness: under diverse visual distractions, our method with video co-training achieves over 70% average success rate.

Problem

• articulated object manipulation
• kinematic structure
• sample efficiency
• data-efficient learning
• robot demonstrations

Innovation

• kinematic-aware interface
• data-efficient manipulation
• articulated object manipulation
• inductive bias
• video co-training