Learning Robot Manipulation from Cross-Morphology Demonstration

TL;DR

This paper introduces MAIL, a framework enabling robots to learn manipulation tasks from demonstrations by agents with different morphologies, effectively bridging the gap in action spaces and handling diverse object properties.

Contribution

MAIL is a novel method that allows learning from demonstrations across significantly different morphologies, including suboptimal and cross-morphology demonstrations, improving robot manipulation capabilities.

Findings

MAIL achieves up to 24% performance improvement over baselines.

It successfully transfers demonstrations from simulated to real robots.

Handles diverse object and cloth properties in manipulation tasks.

Abstract

Some Learning from Demonstrations (LfD) methods handle small mismatches in the action spaces of the teacher and student. Here we address the case where the teacher's morphology is substantially different from that of the student. Our framework, Morphological Adaptation in Imitation Learning (MAIL), bridges this gap allowing us to train an agent from demonstrations by other agents with significantly different morphologies. MAIL learns from suboptimal demonstrations, so long as they provide $\textit{some}$ guidance towards a desired solution. We demonstrate MAIL on manipulation tasks with rigid and deformable objects including 3D cloth manipulation interacting with rigid obstacles. We train a visual control policy for a robot with one end-effector using demonstrations from a simulated agent with two end-effectors. MAIL shows up to $24\%$ improvement in a normalized performance metric over LfD and non-LfD baselines. It is deployed to a real Franka Panda robot, handles multiple variations in properties for objects (size, rotation, translation), and cloth-specific properties (color, thickness, size, material). An overview is on https://uscresl.github.io/mail .