Research On Data-driven Synchronization Of Complex Networks

In recent years,complex networks have been widely used in many fields,such as society,economy,military,biology and so on.The theoretical research of complex networks has attracted the attention of scholars at home and abroad.As one of the important collective behaviors in complex networks,synchronization has become a hot topic.With the continuous development and advancement of science and technology,the dynamics of systems are becoming more and more complex,and it is difficult for us to describe it with accurate mathematical models.This has become a difficult point in the synchronization control of complex networks.Therefore,the research on synchronization of complex networks with unknown dynamics has important practical significance.Recently,data-driven control has been widely used in the control of nonlinear systems with unknown dynamics.Inspired by these,this thesis studies the data-driven synchronization of complex networks.The main contents are as follows:1)The data-driven synchronization of complex networks with unknown dynamic behavior is studied.Firstly,a novel of controller is designed by using pre-compensation technology.Based on the designed controller,an augmented error system is constructed to circumvent the requirements of system dynamics for the control scheme,and the synchronization control is transformed into the optimal performance regulation of the augmented error system with performance function.Secondly,a policy iterative algorithm is proposed to make the iterative performance function converge to the optimal solution of the HJB equation,that is,the optimal performance.Then,a data-driven control scheme is proposed,which is composed of three parts: compensator,controller and critic network.The iterative performance is generated by critic network.On basis of the obtained performance,the compensator is used to construct the iterative control parameter and the controller is used to construct control input.Both the compensator and the critic network are implemented by neural networks.The control scheme only relies on process sampling data,and does not require knowledge of system dynamics.Finally,a robot network is taken as an example to verify the effectiveness of the method.2)The projection synchronization between complex networks with unknown dynamics and different dimensions is studied.Firstly,a synchronization error system is obtained by calculation,and the projection synchronization control is transformed into the optimal performance regulation of the synchronization error system.Secondly,based on the proposed performance function,an event-based projective synchronization control strategy is designed,and Zeno behavior is excluded.Then,the neural network is used to implement the proposed control strategy,and an event triggered data-driven control scheme is proposed.The control scheme consists of identifier,critic network and controller.The identifier is used to estimate the unknown dynamics,the controller is used to construct the optimal control input,and the critic network is used to estimate the optimal performance function.The identifier and critic network are based on neural network.The weights of neural networks and controller are updated only at the eventtriggered instants,otherwise,they will remain unchanged.By using the appropriate eventtriggered threshold and learning rates of neural network,it is proved that the synchronization error asymptotically approaches zero.Due to the introduction of eventtriggered mechanism,the computing burden and communication resources will be saved largely.Finally,the effectiveness of the scheme is verified by a simulation.