Risk-Averse Zero-Order Trajectory Optimization

Marin Vlastelica, Sebastian Blaes, Cristina Pinneri, Georg Martius
Proceedings of the 5th Conference on Robot Learning, PMLR 164:444-454, 2022.

Abstract

We introduce a simple but effective method for managing risk in zero-order trajectory optimization that involves probabilistic safety constraints and balancing of optimism in the face of epistemic uncertainty and pessimism in the face of aleatoric uncertainty of an ensemble of stochastic neural networks. Various experiments indicate that the separation of uncertainties is essential to performing well with data-driven MPC approaches in uncertain and safety-critical control environments.

Cite this Paper

BibTeX
@InProceedings{pmlr-v164-vlastelica22a, title = {Risk-Averse Zero-Order Trajectory Optimization}, author = {Vlastelica, Marin and Blaes, Sebastian and Pinneri, Cristina and Martius, Georg}, booktitle = {Proceedings of the 5th Conference on Robot Learning}, pages = {444--454}, year = {2022}, editor = {Faust, Aleksandra and Hsu, David and Neumann, Gerhard}, volume = {164}, series = {Proceedings of Machine Learning Research}, month = {08--11 Nov}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v164/vlastelica22a/vlastelica22a.pdf}, url = {https://proceedings.mlr.press/v164/vlastelica22a.html}, abstract = {We introduce a simple but effective method for managing risk in zero-order trajectory optimization that involves probabilistic safety constraints and balancing of optimism in the face of epistemic uncertainty and pessimism in the face of aleatoric uncertainty of an ensemble of stochastic neural networks. Various experiments indicate that the separation of uncertainties is essential to performing well with data-driven MPC approaches in uncertain and safety-critical control environments.} }
Endnote
%0 Conference Paper %T Risk-Averse Zero-Order Trajectory Optimization %A Marin Vlastelica %A Sebastian Blaes %A Cristina Pinneri %A Georg Martius %B Proceedings of the 5th Conference on Robot Learning %C Proceedings of Machine Learning Research %D 2022 %E Aleksandra Faust %E David Hsu %E Gerhard Neumann %F pmlr-v164-vlastelica22a %I PMLR %P 444--454 %U https://proceedings.mlr.press/v164/vlastelica22a.html %V 164 %X We introduce a simple but effective method for managing risk in zero-order trajectory optimization that involves probabilistic safety constraints and balancing of optimism in the face of epistemic uncertainty and pessimism in the face of aleatoric uncertainty of an ensemble of stochastic neural networks. Various experiments indicate that the separation of uncertainties is essential to performing well with data-driven MPC approaches in uncertain and safety-critical control environments.
APA
Vlastelica, M., Blaes, S., Pinneri, C. & Martius, G.. (2022). Risk-Averse Zero-Order Trajectory Optimization. Proceedings of the 5th Conference on Robot Learning, in Proceedings of Machine Learning Research 164:444-454 Available from https://proceedings.mlr.press/v164/vlastelica22a.html.

Related Material