Optimal Tracking Control of Linear Discrete-Time Systems Under Cyber Attacks

Abstract

In this paper, the optimal tracking control problem is solved based on the reinforcement learning for linear systems subject to multiple false-data-injection (FDI) attacks. An augmented system is established, which includes the original system and reference-trajectory generator system. The corresponding optimal control issue is formulated as a game problem between the system and malicious adversaries. A Q-learning algorithm is proposed to solve the game algebraic Riccati equation without requiring any knowledge about the dynamics of the augmented system. Finally, an example is provided to show that the system output can track the reference trajectory under cyber attacks.