ICLR 2024
Created by Xingyu Liu, Deepak Pathak and Ding Zhao from Carnegie Mellon University.
If you find our work useful in your research, please cite:
@inproceedings{liu:2024:meta:evolve,
title={Meta-Evolve: Continuous Robot Evolution for One-to-many Policy Transfer},
author={Liu, Xingyu and Pathak, Deepak and Zhao, Ding},
booktitle={International Conference on Learning Representations (ICLR)},
year={2024}
}
We investigate the problem of transferring an expert policy from a source robot to multiple different robots. To solve this problem, we propose a method named Meta-Evolve that uses continuous robot evolution to efficiently transfer the policy to each target robot through a set of tree-structured evolutionary robot sequences. The robot evolution tree allows the robot evolution paths to be shared, so our approach can significantly outperform naive one-to-one policy transfer. We present a heuristic approach to determine an optimized robot evolution tree. Experiments have shown that our method is able to improve the efficiency of one-to-three transfer of manipulation policy by up to 3.2x and one-to-six transfer of agile locomotion policy by 2.4x in terms of simulation cost over the baseline of launching multiple independent one-to-one policy transfers.
Our implementation uses MuJoCo as simulation engine and PyTorch as deep learning framework. The code is tested under Ubuntu 20.04, Python 3.8, mujoco-py 2.1, and PyTorch 2.2.2.
The code and scripts for our Hand Manipulation Suite experiments are in hms/. Please refer to hms/README.md for more details on how to use our code.
The code and scripts for our Hand Manipulation Suite experiments are in dex_ycb/. Please refer to dex_ycb/README.md for more details on how to use our code.
Please refer to LICENSE file.