Stein Variational Model Predictive Control

Alexander Lambert, Fabio Ramos, Byron Boots, Dieter Fox, Adam Fishman
Proceedings of the 2020 Conference on Robot Learning, PMLR 155:1278-1297, 2021.

Abstract

Decision making under uncertainty is critical to real-world, autonomous systems. Model Predictive Control (MPC) methods have demonstrated favorable performance in practice, but remain limited when dealing with complex probability distributions. In this paper, we propose a generalization of MPC that represents a multitude of solutions as posterior distributions. By casting MPC as a Bayesian inference problem, we employ variational methods for posterior computation, naturally encoding the complexity and multi-modality of the decision making problem. We propose a Stein variational gradient descent method to estimate the posterior over control parameters, given a cost function and a sequence of state observations. We show that this framework leads to successful planning in challenging, non-convex optimal control problems.

Cite this Paper

BibTeX
@InProceedings{pmlr-v155-lambert21a, title = {Stein Variational Model Predictive Control}, author = {Lambert, Alexander and Ramos, Fabio and Boots, Byron and Fox, Dieter and Fishman, Adam}, booktitle = {Proceedings of the 2020 Conference on Robot Learning}, pages = {1278--1297}, 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/lambert21a/lambert21a.pdf}, url = {https://proceedings.mlr.press/v155/lambert21a.html}, abstract = {Decision making under uncertainty is critical to real-world, autonomous systems. Model Predictive Control (MPC) methods have demonstrated favorable performance in practice, but remain limited when dealing with complex probability distributions. In this paper, we propose a generalization of MPC that represents a multitude of solutions as posterior distributions. By casting MPC as a Bayesian inference problem, we employ variational methods for posterior computation, naturally encoding the complexity and multi-modality of the decision making problem. We propose a Stein variational gradient descent method to estimate the posterior over control parameters, given a cost function and a sequence of state observations. We show that this framework leads to successful planning in challenging, non-convex optimal control problems.} }
Endnote
%0 Conference Paper %T Stein Variational Model Predictive Control %A Alexander Lambert %A Fabio Ramos %A Byron Boots %A Dieter Fox %A Adam Fishman %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-lambert21a %I PMLR %P 1278--1297 %U https://proceedings.mlr.press/v155/lambert21a.html %V 155 %X Decision making under uncertainty is critical to real-world, autonomous systems. Model Predictive Control (MPC) methods have demonstrated favorable performance in practice, but remain limited when dealing with complex probability distributions. In this paper, we propose a generalization of MPC that represents a multitude of solutions as posterior distributions. By casting MPC as a Bayesian inference problem, we employ variational methods for posterior computation, naturally encoding the complexity and multi-modality of the decision making problem. We propose a Stein variational gradient descent method to estimate the posterior over control parameters, given a cost function and a sequence of state observations. We show that this framework leads to successful planning in challenging, non-convex optimal control problems.
APA
Lambert, A., Ramos, F., Boots, B., Fox, D. & Fishman, A.. (2021). Stein Variational Model Predictive Control. Proceedings of the 2020 Conference on Robot Learning, in Proceedings of Machine Learning Research 155:1278-1297 Available from https://proceedings.mlr.press/v155/lambert21a.html.

Related Material