Model Predictive Control Method For Networked Systems Under Denial Of Service Attacks

Networked control systems,which combine the physical elements of the real world with the computing resources of cyberspace,affect our lives in the future engineering application fields such as industrial automation and intelligence.Compared with the traditional point-topoint control systems,the spatial distributed structure of networked control systems has the advantages of low installation and maintenance cost,flexible reconstruction and so on.However,inserting IT(Internet Technology)components into control systems makes the networked control systems vulnerable to various network attacks.With the increasing connection between networked control systems and many pivotal infrastructures,the risk of network attack also increases greatly.Network security has become a hot issue of social concern.In traditional IT security technology,denial of service(Do S)attack is characterized by the intentional consumption of limited resources such as network bandwidth and packet buffer by a large amount of useless data.In the background of networked control systems,Do S attackers act to prevent the real-time transmission of measurement data or control data,resulting in transmission delay or loss of measurement signal or control signal,which eventually leads to control performance degradation or even instability.Model predictive control(MPC)is an effective method to deal with the impact of Do S attacks.On the one hand,by optimizing a given performance index,MPC can predict a set of optimal control sequences containing current and future control signals to compensate for lost control inputs when packet loss or transmission delay occurs.On the other hand,the development of Ethernet in networked control systems provides the ability to transmit large data packets,which makes the MPC strategy more promising and attractive in practice.From the point of view of controller design,different situations of Do S attacks are analyzed and modeled,and different MPC algorithms are used to ensure that the system can achieve different stability control objectives in the presence of Do S attacks.The main work consists of the following three parts:(1)The stability control problem of the networked systems with actuator constraints and state constraints under Do S attack jamming signals is studied.Do S attacks occur in the communication channel between the controller and the actuator of a single system.First,based on a MPC strategy,a self-triggered mechanism is introduced to balance control performance and communication resources.Second,for the self-triggered condition,the specific MPC design parameters and the maximum continuous step size of Do S attack are given to ensure the exponential stability of a single closed-loop system.(2)The coordinated stability problem of a group of large-scale nonlinear continuous time systems with network information interaction under Do S attacks is studied.Do S attacks occur in the communication link of information exchange between subsystems,and every subsystem has actuator constraints.First,the impact of Do S attack is modeled as a time-varying delay of the subsystem,and the stability of distributed model predictive control(DMPC)strategy with communication delay is considered.Second,based on a robustness constraint and a waiting mechanism,a suitable control update rule is provided,and the feasible range of communication delay,sampling period and cooperation weights among related subsystems caused by Do S attacks are given.The whole distributed system is coordinated and stabilized.Finally,the correctness of the DMPC stability algorithm against Do S attacks is verified on a distributed system composed of four wheeled mobile robots.(3)The consensus control problem of a group of nonlinear continuous time systems with network information interaction under Do S attacks is studied.Do S attacks occur in the communication channel between each subsystem's own sensor and the controller,and every subsystem has actuator constraints.First,in the distributed environment,the Do S attack on each subsystem is modeled as a bounded additive disturbance term from a global perspective.The corresponding model predictive controller of each subsystem iteratively solves a local min-max(minimum cost,maximum attack)optimization problem under the possible worst case(maximum attack).Second,a self-triggered strategy is introduced to reduce the communication burden.Finally,based on the invariant set theory,the sufficient conditions are given to ensure the feasibility of the optimization algorithm and the consensus of the whole distributed system,and the correctness and effectiveness of the self-triggered DMPC consensus algorithm against Do S attacks are verified on multiple car-damper-spring systems.