Resilient Learning-Based Control Under Denial-of-Service Attacks

TL;DR

The paper addresses robust, data-driven output regulation for discrete-time linear systems with unknown parameters subject to Denial-of-Service (DoS) attacks. It introduces a resilient online policy-iteration method that learns the optimal controller from input-state data while an internal-model component preserves stability under DoS, and derives an explicit bound on DoS duration $T^ op$ to guarantee asymptotic tracking. The approach is validated on an inverted pendulum on a cart, illustrating that the learned controller maintains reference tracking despite intermittent communication outages. This work advances cyber-physical resiliency by combining policy iteration with DoS-aware stability analysis in a data-driven, model-free setting.

Abstract

In this paper, we have proposed a resilient reinforcement learning method for discrete-time linear systems with unknown parameters, under denial-of-service (DoS) attacks. The proposed method is based on policy iteration that learns the optimal controller from input-state data amidst DoS attacks. We achieve an upper bound for the DoS duration to ensure closed-loop stability. The resilience of the closed-loop system, when subjected to DoS attacks with the learned controller and an internal model, has been thoroughly examined. The effectiveness of the proposed methodology is demonstrated on an inverted pendulum on a cart.

Figures (3)

• Figure 1: Tracking and disturbance rejection under DoS attacks.
• Figure 2: Convergence of $K_{j}$ to $K^\star$.
• Figure 3: Convergence of $P_{j}$ to $P^\star$.

Theorems & Definitions (10)

• Definition II.1
• Remark II.1
• Lemma II.1
• proof
• Theorem III.1
• proof
• Lemma III.1
• proof
• Remark IV.1
• Remark IV.2