X-Imitator: Spatial-Aware Imitation Learning via Bidirectional Action-Pose Interaction

TL;DR

X-Imitator introduces a bidirectional, modular framework for robotic manipulation that tightly couples spatial perception and action generation, enabling continuous mutual refinement and outperforming prior methods.

Contribution

It proposes a novel dual-path, bidirectional architecture that models spatial perception and action as a coupled loop, mimicking human internal forward models.

Findings

Outperforms vanilla policies and prior pose-guided methods in 24 simulated and 3 real-world tasks.

Enables continuous mutual refinement between spatial reasoning and action generation.

Designed as a modular system that can be integrated into various visuomotor policies.

Abstract

Effectively handling the interplay between spatial perception and action generation remains a critical bottleneck in robotic manipulation. Existing methods typically treat spatial perception and action execution as decoupled or strictly unidirectional processes, fundamentally restricting a robot's ability to master complex manipulation tasks. To address this, we propose X-Imitator, a versatile dual-path framework that models spatial perception and action execution as a tightly coupled bidirectional loop. By reciprocally conditioning current pose predictions on past actions and vice versa, this framework enables continuous mutual refinement between spatial reasoning and action generation. This joint modeling exactly mimics human internal forward models. Designed as a modular architecture, the system can be seamlessly integrated into various visuomotor policies. Extensive experiments across 24 simulated and 3 real-world tasks demonstrate that our framework significantly outperforms both vanilla policies and prior methods utilizing explicit pose guidance. The code will be open sourced.