Safe Learning MPC with Limited Model Knowledge and Data

arXiv2004.00759MaRDI QIDQ6337893

Author name not available (Why is that?)

Publication date: 1 April 2020

Abstract: This paper presents an end-to-end framework for safe learning-based control (LbC) using nonlinear stochastic MPC and distributionally robust optimization (DRO). This work is motivated by several open challenges in LbC literature. In particular, many control-theoretic LbC methods require subject matter expertise in order to translate their own safety guarantees, often manifested as preexisting data of safe trajectories or structural model knowledge. In this paper, we focus on LbC where the controller is applied directly to a system of which it has no or extremely limited direct experience, towards safety during tabula-rasa or ``blank slate model-based learning and control. This explores the boundary of the status-quo in control theory relating to requirements for subject matter expertise. We show under basic and limited assumptions on the underlying problem, we can translate probabilistic guarantees on feasibility to nonlinear systems using existing results in stochastic MPC and DRO literature. We also present a coupled and intuitive formulation for persistence of excitation (PoE), and illustrate the connection between PoE and applicability of the proposed method. Our case studies of vehicle obstacle avoidance and safe extreme fast charging of lithium-ion batteries reveal powerful empirical results supporting the underlying DRO theory. Our method is widely applicable within the LbC domain to applications including for example airborne wind energy systems, vehicle obstacle avoidance, energy storage systems management, and video compression.

Has companion code repository: https://github.com/aaronkandel/wasserstein-nonlinear-mpc

This page was built for publication: Safe Learning MPC with Limited Model Knowledge and Data

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q6337893)