Safety Filter Design for Neural Network Systems via Convex Optimization
Abstract
With the increase in data availability, it has been widely demonstrated that neural networks (NN) can capture complex system dynamics precisely in a data-driven manner. However, the architectural complexity and nonlinearity of the NNs make it challenging to synthesize a provably safe controller. In this work, we propose a novel safety filter that relies on convex optimization to ensure safety for a NN system, subject to additive disturbances that are capable of capturing modeling errors. Our approach leverages tools from NN verification to over-approximate NN dynamics with a set of linear bounds, followed by an application of robust linear MPC to search for controllers that can guarantee robust constraint satisfaction. We demonstrate the efficacy of the proposed framework numerically on a nonlinear pendulum system.
Cite
@article{arxiv.2308.08086,
title = {Safety Filter Design for Neural Network Systems via Convex Optimization},
author = {Shaoru Chen and Kong Yao Chee and Nikolai Matni and M. Ani Hsieh and George J. Pappas},
journal= {arXiv preprint arXiv:2308.08086},
year = {2023}
}
Comments
This paper has been accepted to the 2023 62nd IEEE Conference on Decision and Control (CDC)