Deep Kernels for Optimizing Locomotion Controllers

Rika Antonova, Akshara Rai, Christopher G. Atkeson
Proceedings of the 1st Annual Conference on Robot Learning, PMLR 78:47-56, 2017.

Abstract

Sample efficiency is important when optimizing parameters of locomotion controllers, since hardware experiments are time consuming and expensive. Bayesian Optimization, a sample-efficient optimization framework, has recently been widely applied to address this problem, but further improvements in sample efficiency are needed for practical applicability to real-world robots and high-dimensional controllers. To address this, prior work has proposed using domain expertise for constructing custom distance metrics for locomotion. In this work we show how to learn such a distance metric automatically. We use a neural network to learn an informed distance metric from data obtained in high-fidelity simulations. We conduct experiments on two different controllers and robot architectures. First, we demonstrate improvement in sample efficiency when optimizing a 5-dimensional controller on the ATRIAS robot hardware. We then conduct simulation experiments to optimize a 16-dimensional controller for a 7-link robot model and obtain significant improvements even when optimizing in perturbed environments. This demonstrates that our approach is able to enhance sample efficiency for two different controllers, hence is a fitting candidate for further experiments on hardware in the future.

Cite this Paper


BibTeX
@InProceedings{pmlr-v78-antonova17a, title = {Deep Kernels for Optimizing Locomotion Controllers}, author = {Antonova, Rika and Rai, Akshara and Atkeson, Christopher G.}, booktitle = {Proceedings of the 1st Annual Conference on Robot Learning}, pages = {47--56}, year = {2017}, editor = {Levine, Sergey and Vanhoucke, Vincent and Goldberg, Ken}, volume = {78}, series = {Proceedings of Machine Learning Research}, month = {13--15 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v78/antonova17a/antonova17a.pdf}, url = {https://proceedings.mlr.press/v78/antonova17a.html}, abstract = {Sample efficiency is important when optimizing parameters of locomotion controllers, since hardware experiments are time consuming and expensive. Bayesian Optimization, a sample-efficient optimization framework, has recently been widely applied to address this problem, but further improvements in sample efficiency are needed for practical applicability to real-world robots and high-dimensional controllers. To address this, prior work has proposed using domain expertise for constructing custom distance metrics for locomotion. In this work we show how to learn such a distance metric automatically. We use a neural network to learn an informed distance metric from data obtained in high-fidelity simulations. We conduct experiments on two different controllers and robot architectures. First, we demonstrate improvement in sample efficiency when optimizing a 5-dimensional controller on the ATRIAS robot hardware. We then conduct simulation experiments to optimize a 16-dimensional controller for a 7-link robot model and obtain significant improvements even when optimizing in perturbed environments. This demonstrates that our approach is able to enhance sample efficiency for two different controllers, hence is a fitting candidate for further experiments on hardware in the future.} }
Endnote
%0 Conference Paper %T Deep Kernels for Optimizing Locomotion Controllers %A Rika Antonova %A Akshara Rai %A Christopher G. Atkeson %B Proceedings of the 1st Annual Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2017 %E Sergey Levine %E Vincent Vanhoucke %E Ken Goldberg %F pmlr-v78-antonova17a %I PMLR %P 47--56 %U https://proceedings.mlr.press/v78/antonova17a.html %V 78 %X Sample efficiency is important when optimizing parameters of locomotion controllers, since hardware experiments are time consuming and expensive. Bayesian Optimization, a sample-efficient optimization framework, has recently been widely applied to address this problem, but further improvements in sample efficiency are needed for practical applicability to real-world robots and high-dimensional controllers. To address this, prior work has proposed using domain expertise for constructing custom distance metrics for locomotion. In this work we show how to learn such a distance metric automatically. We use a neural network to learn an informed distance metric from data obtained in high-fidelity simulations. We conduct experiments on two different controllers and robot architectures. First, we demonstrate improvement in sample efficiency when optimizing a 5-dimensional controller on the ATRIAS robot hardware. We then conduct simulation experiments to optimize a 16-dimensional controller for a 7-link robot model and obtain significant improvements even when optimizing in perturbed environments. This demonstrates that our approach is able to enhance sample efficiency for two different controllers, hence is a fitting candidate for further experiments on hardware in the future.
APA
Antonova, R., Rai, A. & Atkeson, C.G.. (2017). Deep Kernels for Optimizing Locomotion Controllers. Proceedings of the 1st Annual Conference on Robot Learning, in Proceedings of Machine Learning Research 78:47-56 Available from https://proceedings.mlr.press/v78/antonova17a.html.

Related Material