Reinforcement Learning-Based Sliding Mode Tracking Control for the Two-Time-Scale Systems: Dealing With Actuator Attacks
Ying Zhang, Lei Ma, Chunyu Yang, Wei Dai
Abstract
This brief focuses on the security sliding mode tracking control (SMTC) problem for the two-time-scale systems (TTSs) subjected to actuator attacks. All the system parameters are supposed to be completely unavailable to the defenders, and therefore a novel <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\varepsilon $ </tex-math></inline-formula> -dependent model-free sliding mode function is designed to cope with such a situation. Then, a new SMTC law, consisting of an optimal tracking controller and an attack compensator, is developed to guarantee the tracking performance of TTSs under actuator attacks. Furthermore, a data-driven SMTC strategy is constructed in a model-free approach. More specifically, the developed SMTC algorithm is implemented by the reinforcement learning method by using the slow and fast sampled data simultaneously. It should be pointed out that both the ill-conditioned numerical issue caused by the coupling of slow-fast dynamics is avoided and the actuator attacks are attenuated. Finally, a networked DC motor example demonstrates the proposed algorithm.