Research On Model Predictive Control Theory Based On Dynamic Event-Triggered Under Networked Attacks

Networked Control System(NCS)is a kind of distributed control system with deep integration of cyber space and physical space,in which the data interaction can be realized via communication networks among controller,local sensors and actuators.This control method solves the wired complexity to a great extent,and realizes resource sharing and remote control.However,the introduction of network makes the system exposed to the insecure networked environment and threatened by various network problems,among which the most prominent problem is network constraints and cyber attacks.The traditional predictive control algorithm cannot solve those network problems effectively.If it handled improperly,the limited communication resources and bandwidth will be wasted in the view of communication resources utilization,and the system performance will deteriorate or even become unstable in the view of system operation.Therefore,extensive attention is paid to design an appropriate control algorithm,which balances the control performance level and stability level.Based on above,this paper studies the discrete linear NCS with external disturbance,DoS attacks,deception attacks,and parameter uncertainty,aiming to design a class of effective Model Predictive Control(MPC)algorithm to reduce the negative impact caused by networked unreliability.The main research contents of this paper are as follows:(1)Considering the linear discrete NCS with DoS attack and external interference,an MPC algorithm based on dynamic event-triggered mechanism is proposed.Firstly,to solve the communication constraint caused by bandwidth limitation,the dynamic event triggering mechanism is introduced while designing the system model.Secondly,Markov model and probability transfer are used to describe DoS attacks.For the controller subjects to DoS attacks and external interference,the concept of mean square input-state stability(ISS)is introduced to designed the controller.Then,the model-dependent Lyapunov function is constructed,and the performance upper bound conditions,invariant set conditions and constraints are derived in the form of LMIs.On this basis,the MPC problem is transformed into a convex optimization problem with LMI constraints.Finally,the recursion feasibility of the MPC algorithm and the closed-loop mean-square ISS is discussed.(2)Considering the linear discrete NCS with external interference,deception attacks and communication constraints,an MPC algorithm is studied to improve the robustness of the system.Firstly,in the situation of communication constraints caused by limited bandwidth,the event triggering mechanism and stochastic communication protocol are introduced into the system model design to reduce the communication burden.The former can decrease the frequency of data transmission,while the latter can reduce the amount of data transmitted at each instant.Secondly,in the situation of controller subjected to the deception attacks and external interference,the concept of mean-square ISS is introduced into the design of MPC controller,and a set of feedback gain and the event-triggered matrix in the framework of MPC are determined,so that the closed-loop system is mean-square ISS with a certain level of performance.Thirdly,a protocol-dependent Lyapunov function is constructed,and a sufficient condition for the existence of an ideal MPC controller is derived,which is expressed as a convex optimization problem with linear matrix inequalities.Finally,the recursive feasibility of the MPC algorithm is analyzed by the closed-loop mean-square error.