Sample-efficient Cross-Entropy Method for Real-time Planning

Cristina Pinneri, Shambhuraj Sawant, Sebastian Blaes, Jan Achterhold, Joerg Stueckler, Michal Rolinek, Georg Martius
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:1049-1065, 2021.

Abstract

Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-pinneri21a, title = {Sample-efficient Cross-Entropy Method for Real-time Planning}, author = {Pinneri, Cristina and Sawant, Shambhuraj and Blaes, Sebastian and Achterhold, Jan and Stueckler, Joerg and Rolinek, Michal and Martius, Georg}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {1049--1065}, year = {2021}, editor = {Kober, Jens and Ramos, Fabio and Tomlin, Claire}, volume = {155}, series = {Proceedings of Machine Learning Research}, month = {16--18 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v155/pinneri21a/pinneri21a.pdf}, url = {https://proceedings.mlr.press/v155/pinneri21a.html}, abstract = {Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.} }
Endnote
%0 Conference Paper %T Sample-efficient Cross-Entropy Method for Real-time Planning %A Cristina Pinneri %A Shambhuraj Sawant %A Sebastian Blaes %A Jan Achterhold %A Joerg Stueckler %A Michal Rolinek %A Georg Martius %B Proceedings of the 2020 Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2021 %E Jens Kober %E Fabio Ramos %E Claire Tomlin %F pmlr-v155-pinneri21a %I PMLR %P 1049--1065 %U https://proceedings.mlr.press/v155/pinneri21a.html %V 155 %X Trajectory optimizers for model-based reinforcement learning, such as the Cross-Entropy Method (CEM), can yield compelling results even in high-dimensional control tasks and sparse-reward environments. However, their sampling inefficiency prevents them from being used for real-time planning and control. We propose an improved version of the CEM algorithm for fast planning, with novel additions including temporally-correlated actions and memory, requiring 2.7-22x less samples and yielding a performance increase of 1.2-10x in high-dimensional control problems.
APA
Pinneri, C., Sawant, S., Blaes, S., Achterhold, J., Stueckler, J., Rolinek, M. & Martius, G.. (2021). Sample-efficient Cross-Entropy Method for Real-time Planning. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:1049-1065 Available from https://proceedings.mlr.press/v155/pinneri21a.html.

Related Material