Movement Primitive Diffusion: Learning Gentle Robotic Manipulation of Deformable Objects

TL;DR

This paper introduces Movement Primitive Diffusion (MPD), a novel imitation learning method that enables gentle and data-efficient manipulation of deformable objects in robotic surgery, combining diffusion-based IL with probabilistic movement primitives.

Contribution

MPD is the first method to integrate diffusion-based imitation learning with probabilistic movement primitives for delicate deformable object manipulation in RAS.

Findings

MPD outperforms state-of-the-art methods in success rate.

MPD achieves higher motion quality.

MPD demonstrates superior data efficiency.

Abstract

Policy learning in robot-assisted surgery (RAS) lacks data efficient and versatile methods that exhibit the desired motion quality for delicate surgical interventions. To this end, we introduce Movement Primitive Diffusion (MPD), a novel method for imitation learning (IL) in RAS that focuses on gentle manipulation of deformable objects. The approach combines the versatility of diffusion-based imitation learning (DIL) with the high-quality motion generation capabilities of Probabilistic Dynamic Movement Primitives (ProDMPs). This combination enables MPD to achieve gentle manipulation of deformable objects, while maintaining data efficiency critical for RAS applications where demonstration data is scarce. We evaluate MPD across various simulated and real world robotic tasks on both state and image observations. MPD outperforms state-of-the-art DIL methods in success rate, motion quality, and data efficiency. Project page: https://scheiklp.github.io/movement-primitive-diffusion/