Diff-Transfer: Model-based Robotic Manipulation Skill Transfer via Differentiable Physics Simulation

TL;DR

Diff-Transfer introduces a framework that uses differentiable physics simulation to efficiently transfer robotic manipulation skills by discovering task paths and adapting actions through gradient-guided sub-task transitions.

Contribution

It presents a novel method combining path planning and differentiable physics for skill transfer in robotic manipulation, enabling efficient adaptation across tasks.

Findings

Successfully transfers skills in simulation experiments

Demonstrates effectiveness on four challenging transfer tasks

Utilizes gradient information for action adaptation

Abstract

The capability to transfer mastered skills to accomplish a range of similar yet novel tasks is crucial for intelligent robots. In this work, we introduce $\textit{Diff-Transfer}$, a novel framework leveraging differentiable physics simulation to efficiently transfer robotic skills. Specifically, $\textit{Diff-Transfer}$ discovers a feasible path within the task space that brings the source task to the target task. At each pair of adjacent points along this task path, which is two sub-tasks, $\textit{Diff-Transfer}$ adapts known actions from one sub-task to tackle the other sub-task successfully. The adaptation is guided by the gradient information from differentiable physics simulations. We propose a novel path-planning method to generate sub-tasks, leveraging $Q$-learning with a task-level state and reward. We implement our framework in simulation experiments and execute four challenging transfer tasks on robotic manipulation, demonstrating the efficacy of $\textit{Diff-Transfer}$ through comprehensive experiments. Supplementary and Videos are on the website https://sites.google.com/view/difftransfer