Generating Adversarial Disturbances for Controller Verification
This work provides a method for generating adversarial disturbances, which is crucial for verifying the robustness of controllers, particularly for safety-critical systems. It is an incremental improvement on existing methods for controller verification.
This paper addresses the challenge of generating maximally adversarial disturbances for blackbox controllers. The authors propose an online learning approach, MOTR, which adaptively generates disturbances to maximize the cost incurred by the controller. The method is shown to outperform baseline approaches like H-infinity disturbance generation and gradient-based methods on linear systems and wind disturbances for the PX4 controller in AirSim.
We consider the problem of generating maximally adversarial disturbances for a given controller assuming only blackbox access to it. We propose an online learning approach to this problem that \emph{adaptively} generates disturbances based on control inputs chosen by the controller. The goal of the disturbance generator is to minimize \emph{regret} versus a benchmark disturbance-generating policy class, i.e., to maximize the cost incurred by the controller as well as possible compared to the best possible disturbance generator \emph{in hindsight} (chosen from a benchmark policy class). In the setting where the dynamics are linear and the costs are quadratic, we formulate our problem as an online trust region (OTR) problem with memory and present a new online learning algorithm (\emph{MOTR}) for this problem. We prove that this method competes with the best disturbance generator in hindsight (chosen from a rich class of benchmark policies that includes linear-dynamical disturbance generating policies). We demonstrate our approach on two simulated examples: (i) synthetically generated linear systems, and (ii) generating wind disturbances for the popular PX4 controller in the AirSim simulator. On these examples, we demonstrate that our approach outperforms several baseline approaches, including $H_{\infty}$ disturbance generation and gradient-based methods.