In this section
@inproceedings{Lim:2022:10.1109/ICRA46639.2022.9811559,
author = {Lim, BWT and Grillotti, L and Bernasconi, L and Cully, A},
doi = {10.1109/ICRA46639.2022.9811559},
pages = {5360--5366},
publisher = {IEEE},
title = {Dynamics-aware quality-diversity for efficient learning of skill repertoires},
url = {http://dx.doi.org/10.1109/ICRA46639.2022.9811559},
year = {2022}
}
TY - CPAPER
AB - Quality-Diversity (QD) algorithms are powerful exploration algorithms that allow robots to discover large repertoires of diverse and high-performing skills. However, QD algorithms are sample inefficient and require millionsof evaluations. In this paper, we propose Dynamics-Aware Quality-Diversity (DA-QD), a framework to improve the sample efficiency of QD algorithms through the use of dynamics models. We also show how DA-QD can then be used for continual acquisition of new skill repertoires. To do so, weincrementally train a deep dynamics model from experience obtained when performing skill discovery using QD. We can then perform QD exploration in imagination with an imagined skill repertoire. We evaluate our approach on three robotic experiments. First, our experiments show DA-QD is 20 timesmore sample efficient than existing QD approaches for skill discovery. Second, we demonstrate learning an entirely new skill repertoire in imagination to perform zero-shot learning. Finally, we show how DA-QD is useful and effective for solving a long horizon navigation task and for damage adaptation in the real world. Videos and source code are available at: https://sites.google.com/view/da-qd.
AU - Lim,BWT
AU - Grillotti,L
AU - Bernasconi,L
AU - Cully,A
DO - 10.1109/ICRA46639.2022.9811559
EP - 5366
PB - IEEE
PY - 2022///
SP - 5360
TI - Dynamics-aware quality-diversity for efficient learning of skill repertoires
UR - http://dx.doi.org/10.1109/ICRA46639.2022.9811559
UR - http://ieeexplore.ieee.org/abstract/document/9811559
UR - http://hdl.handle.net/10044/1/96112
ER -
