TacDiffusion: Force-domain Diffusion Policy for Precise Tactile Manipulation

2024-09-17

arXiv.org

AI Summary

To address the challenge of zero-shot skill transfer for high-precision haptic assembly across unseen workpieces, this paper proposes Force-Domain Diffusion—a novel approach that directly generates control commands in the 6D wrench space. Methodologically, it introduces the first diffusion-based paradigm with wrenches as outputs; designs a dynamic-system filter to bridge diffusion inference with real-time control frequency; and provides a practical deployment guideline for trading off inference speed against accuracy. Evaluated on unseen high-precision insertion tasks using only single-task demonstrations, the method achieves a 95.7% zero-shot transfer success rate—outperforming baseline methods by 9.15 percentage points. This demonstrates substantial improvements in generalization capability and engineering deployability for contact-rich robotic manipulation.

Technology Category

• Intelligent Robots: Manipulation
• Computer Vision: Diffusion Models for Vision
• Humans and AI: Human-Aware Planning and Behavior Prediction

Application Category

• Economics, Online Markets and Human Computation: Trust and reliance of crowd workers and data experts on GenAI
• Systems and Infrastructure for Web, Mobile and WoT: Experiences and lessons learnt from Web-based algorithms and system deployments
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Assembly is a crucial skill for robots in both modern manufacturing and service robotics. However, mastering transferable insertion skills that can handle a variety of high-precision assembly tasks remains a significant challenge. This paper presents a novel framework that utilizes diffusion models to generate 6D wrench for high-precision tactile robotic insertion tasks. It learns from demonstrations performed on a single task and achieves a zero-shot transfer success rate of 95.7% across various novel high-precision tasks. Our method effectively inherits the self-adaptability demonstrated by our previous work. In this framework, we address the frequency misalignment between the diffusion policy and the real-time control loop with a dynamic system-based filter, significantly improving the task success rate by 9.15%. Furthermore, we provide a practical guideline regarding the trade-off between diffusion models' inference ability and speed.

Problem

• Develops a framework for high-precision tactile robotic insertion tasks.
• Achieves zero-shot transfer across various high-precision assembly tasks.
• Addresses frequency misalignment to improve task success rate.

Innovation

• Uses diffusion models for 6D wrench generation
• Implements dynamic system-based filter for frequency alignment
• Achieves 95.7% zero-shot transfer success rate