Reinforcement Learning Based Fault-Alarmed Hybrid Control Design for Nonlinear Cyber-Physical Systems via Negative Determination Lemma

This article delineates the stabilization problem of non-linear cyber physical systems by integrating the reinforcement learning based switched fault alarm controller and quadratic function negative determination lemma. To begin with, quadratic function based negative determination lemma is applied to handle the input delay occurred in the controller which derives the sufficient conditions for the quadratic polynomials arises in the system by partitioning the delay-intervals. Subsequently, a robust reinforcement learning control is designed in conjunction with a switched fault alarm. In particular, the alarm threshold value is used to determine the fault alarmed signal, which is intended to promptly notify the controller to jump from the robust learning controller to the fault-tolerant controller. Based on the Lyapunov stability theory and linear matrix inequality approach, the asymptotic stability criterion is developed to sustain skillfully in the presence of actuator-faults in the system as well as attains the accurate convergence with the aid of reinforced learning controller. Further, numerical simulations are performed to validate the proposed approach.