Anti-disturbance And Security Control Of A Class Of Nontriangular Nonlinear System Based On Predictor

Owing to more general system models and the broad application scenarios such as the continuous stirred tank reactor,multi-quadrotor unmanned aerial vehicles,nontriangular nonlinear systems and multi-agent systems have attracted much attention in the control community.Due to the complex situations such as unknown disturbances and cyberattacks,the control performance of nontriangular systems may be degraded,and even the system may be unstable.In this paper,considering the complex situations such as unknown external disturbance and cyberattacks,constructing predictor to predict the state of the system and improve the transient performance of neural network learning behavior,through disturbance observer,attack compensator,state observer and other technologies,a series of studies are carried out on the anti-disturbance and security control of nontriangular nonlinear systems and multi-agent systems under cyberattacks under the framework of backstepping method.The main research work includes the three aspects:1)For a class of nontriangular nonlinear systems in presence of unknown disturbances,we propose a predictor-based neural dynamic surface control(PNDSC)strategy in this paper.This nontriangular system is transformed via the mean value theorem,and a predictor is then constructed based on the transformed model to predict the system state.The PNDSC in this paper utilizes the difference between the prediction state and the actual state of the system to update learning parameters for improving neural networks’(NNs)transient learning behaviors with overlarge adaptive gains.On the basis of improved NNs’ approximation behaviors,a predictor-based NNs disturbance observer is constructed for compensation for external disturbances and approximation errors from NNs.Furthermore,with predictors,a normalization method of weights is developed to reduce the number of online learning parameters.To avoid an algebraic loop problem,partial state vectors are employed as input signals of NNs for approximating unknown dynamics,and compensation items are designed to compensate for approximation errors from NNs.On the basis of the aforementioned result,distributed output consensus control for a class of nontriangular multi-agent systems under measurement noises are taken into account.Because of the existence of measurement noise,the actual state information of the system can not be obtained.We construct predictor to predict the system state,and use predictor states to replace the measurement states containing noise to construct the surface error,and the predictor state is used as the input vector of the neural network to design the control strategy,so as to avoid the direct influence of measurement noise on the control performance.At the same time,according to the topology knowledge of graph theory,the distributed output consensus of multi-agent system under external disturbance and measurement noise is realized by communicating with local neighbors.By constructing Lyapunov equation,it is proved that the closed-loop system signal is bounded under the condition of bounded initial state.2)This paper addresses a secure predictor-based neural dynamic surface control issue for a cyber-physical system in a nontriangular form deception attacks.With introduction of nonlinear gain functions into the controller,instability of the system is avoided caused by overlarge control input due to the sudden change of the error signals when system suffers from deception attacks unexpectedly.We utilize neural networks to estimate the upper bound of the actuator attacks and compensate their adverse impacts on the system.On the other hand,by constructing intermediate variables with the help of predictor,the sensor attack compensator is updated to deal with the sensor deception attacks.On this basis,the above control strategy is applied to the containment control problem of nontriangular multi-agents systems,and the influence of external disturbance and input dead zone on the system is also considered.The effects of actuator attack are taken as a shift in dead-zone nonlinearity and is defined as generalized disturbance together with external disturbance and approximation errors from NNs.A disturbance observer is developed with predictors to compensate the effect of the generalized disturbances.By a Lyapunov function,it is proved that the closed-loop signal is bounded when the initial state of the system is bounded.3)This paper presents a security predictor-based adaptive distributed output consensus control strategy for a class of nontriangular multi-agents systems subjected to denial-of-service(Do S)attacks.To solve the issue that all the state variables are unavailable when systems suffer from Do S attacks,this paper constructs a switched-type adaptive state estimator to estimate its state.On the basis,a predictor is introduced.Since predictor states are used instead of switching between the actual state and the observer state,the control strategy proposed in this paper can avoid the high-frequency oscillation in the neural network learning behavior under excessive adaptive gain.This paper introduces nonlinear gain functions into control strategy,so as to avoid the problem of excessive control input caused by sudden change of error signals when Do S attack is active or sleeping.At the same time,in order to realize the distributed output consensus of nontriangular multi-agent systems under Do S attack,a switched-type adaptive state estimator is designed on the agent to observe the information of neighbor agents.When the Do S attack is active in the communication link,by using the observed information of the neighbor agents,the control strategy is designed to ensure that the output consensus of systems even when the communication link is destroyed.This paper removes the assumption that the communication topology keep connected under DOS attack,and expands the application scope of the control strategy.By a Lyapunov function,it is proved that the proposed control strategy can ensure that all signals in the system are bounded and the convergence of tracking error arrive in a small neighborhood around the origin.