File size: 1,702 Bytes
25cd4b9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 |
---
license: mit
language:
- en
tags:
- robotics
- robotic-manipulation
- deformable-object
- cross-morphology-demonstration
task_categories:
- robotics
---
# Learning Robot Manipulation from Cross-Morphology Demonstration (CoRL 2023)
[[Project website](https://uscresl.github.io/mail/)]
[[arXiv PDF](https://arxiv.org/abs/2304.03833)]
Datasets for [MAIL](https://github.com/uscresl/mail).
Authors: [Gautam Salhotra](https://www.gautamsalhotra.com/)\*, [I-Chun Arthur Liu](https://arthurliu.com/)\*, [Gaurav S. Sukhatme](https://robotics.usc.edu/~gaurav/) (* denotes equal contribution)
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 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). |