Diffusion Policy for Coordinated Control of a Nonholonomic Mobile Base and Dual Arms in Door Opening and Passing

2026-05-14

AI Summary

Existing robotic systems struggle to coordinate dual arms and a nonholonomically constrained mobile base in a human-like manner to perform the demanding task of pulling and traversing through heavy self-closing doors. This work proposes an end-to-end diffusion policy that directly generates long-horizon, multimodal action sequences from visual inputs, enabling tight coordination between the mobile base and dual arms throughout the complete door-opening and traversal task—without relying on conventional state machines. To the best of our knowledge, this is the first approach to apply a single diffusion model to such a high-dimensional, nonholonomic system. The method significantly improves task success rates and demonstrates strong robustness to environmental disturbances, particularly in high-damping door-pulling scenarios.

Abstract

Opening heavy, self closing doors, especially those that require pulling remains a long standing challenge in robotics. Humans naturally employ both arms in a dexterous manner, rotating the handle, widening the gap, holding the door, switching arms when needed, and moving through while maintaining clearance. To replicate such behaviors, a robot must perform a long sequence of motions spanning multiple stages and interactions with different parts of the door. Traditional approaches rely on state machines that transition between manually defined stages (e.g., pulling after the knob is rotated, passing after the gap is sufficiently wide). While intuitive, these methods lack robustness, as hand crafted trajectories fail to generalize to the diversity of real world conditions without extensive engineering effort. Recent advances in imitation learning offer a scalable alternative, yet no existing visual action model has demonstrated simultaneous coordination of a nonholonomic base and dual arms for the complete door opening and passing task. In this paper, we tackle this complex, highly constrained problem using a diffusion based visuomotor control policy. Our results demonstrate that a single end to end policy can be learned to execute long horizon tasks requiring tight coordination between manipulation and locomotion. The resulting policy not only achieves a high success rate in opening and traversing damped pull doors but also demonstrates strong robustness to external disturbances capabilities that are difficult to realize with traditional methods.

Problem

• door opening
• nonholonomic mobile base
• dual-arm coordination
• visuomotor control
• long-horizon manipulation

Innovation

• diffusion policy
• visuomotor control
• nonholonomic mobile base
• dual-arm coordination
• door opening and passing