Adversarial Model Predictive Control via Second-Order Cone Programming

TL;DR

This paper introduces a polynomial-time method for designing optimal adversarial attacks on safety-critical systems within a model predictive control framework, focusing on positive systems like autonomous vehicles and microgrids.

Contribution

It characterizes a class of NP-hard attack problems solvable via second-order cone programming, specifically applying to positive systems.

Findings

Optimal attacks on autonomous vehicle systems demonstrated.

Microgrid attack strategies successfully formulated.

Polynomial-time solutions for a class of attack problems.

Abstract

We study the problem of designing attacks to safety-critical systems in which the adversary seeks to maximize the overall system cost within a model predictive control framework. Although in general this problem is NP-hard, we characterize a family of problems that can be solved in polynomial time via a second-order cone programming relaxation. In particular, we show that positive systems fall under this family. We provide examples demonstrating the design of optimal attacks on an autonomous vehicle and a microgrid.