Residual Robot Learning for Object-Centric Probabilistic Movement Primitives

TL;DR

This paper introduces a method combining Probabilistic Movement Primitives with Residual Reinforcement Learning to improve the precision of robotic object manipulation tasks, demonstrated on a 3D block insertion task with a 7-DoF robot.

Contribution

It presents a novel approach that integrates residual RL with ProMPs, enhancing manipulation accuracy in robotic tasks beyond traditional ProMP capabilities.

Findings

Successful learning of a complex insertion task

Enhanced precision over basic ProMPs

Effective use of demonstration variability as a decision variable

Abstract

It is desirable for future robots to quickly learn new tasks and adapt learned skills to constantly changing environments. To this end, Probabilistic Movement Primitives (ProMPs) have shown to be a promising framework to learn generalizable trajectory generators from distributions over demonstrated trajectories. However, in practical applications that require high precision in the manipulation of objects, the accuracy of ProMPs is often insufficient, in particular when they are learned in cartesian space from external observations and executed with limited controller gains. Therefore, we propose to combine ProMPs with recently introduced Residual Reinforcement Learning (RRL), to account for both, corrections in position and orientation during task execution. In particular, we learn a residual on top of a nominal ProMP trajectory with Soft-Actor Critic and incorporate the variability in the demonstrations as a decision variable to reduce the search space for RRL. As a proof of concept, we evaluate our proposed method on a 3D block insertion task with a 7-DoF Franka Emika Panda robot. Experimental results show that the robot successfully learns to complete the insertion which was not possible before with using basic ProMPs.