Consensus Tracking Iterative Learning Control For Multi-agent Systems

Through cooperation and coordination,multiple agents can greatly improve the performance of the system.The consensus of multi-agent systems means that all agents only depend on local information to achieve global control according to the given topological relationship,and finally make some states of all agents converge.For the multi-agent systems with repetitive motion,iterative learning control(ILC)has been paid great attention to achieve consensus control of the system.The design of ILC is using the previous system information to structure the control input such that we can improve the system performance gradually along the iteration axis.However,due to the influence of the system itself or external factors,the trial lengths may be different from the expected length.This is the ILC with varying trial lengths.This paper studies the ILC problem of multi-agent systems under complex situations and ILC with varying trial lengths.Main contributions are as follows:1.Chapter 3 considers the discrete-time linear multi-agent systems with random noises and measurement range limitations.In this chapter,the case of measurement noises and communication noises is considered,a decreasing sequence is introduced to reduce the impact of random noises,and a random variable is used to connect the original data with the measured data.The convergence of the system is proved by the stochastic approximation technique.The results are extended to the switching topology case,and the effectiveness of the method is verified by simulations.2.Chapter 4 considers the continuous-time nonlinear multi-agent systems with input disturbance and unknown input gain.The neural networks used as an approximator is designed to compensate for the system's nonlinearity.The robust learning component is used to overcome the unknown input gain and input disturbance.An adaptive law combining time-and iteration-domain is used to tune the controller parameters.Finally,the convergence of tracking error is proved by constructing an appropriate composite energy function,and the effectiveness of the control law is verified by simulation.3.Chapter 5 considers the continuous-time nonlinear multi-agent systems with varying trial lengths.In this chapter,the tracking error compensation mechanism is introduced into the controller design.Then we design the adaptive ILC law and the parameter adaptive updating law.Finally,the convergence of the system error is proved and the effectiveness of the control law is verified by simulation.Varying trial lengths is one of the important problems in ILC.In order to better to solve the varying trial lengths problem of the multi-agent systems,we study the information compensation mechanism and the design of the controller of a single system.The main results are as follows:4.Chapter 6 studies the iteration varying trail lengths problem for high-order continuous-time nonlinear systems,where the initial state may deviate from the desired value and the sign of input gain is unknown.The neural networks and a Nussbaum function are used to cope with the system's nonlinearity and unknown control direction,respectively.Furthermore,the initial state deviation issue is resolved by introducing an initial state learning protocol.Finally,based on the virtual tracking error,a composite energy function is constructed to show the effectiveness of the proposed algorithm.5.Chapter 7 studies the tracking problem of discrete time-varying linear systems with randomly varying trial lengths.Using the two-dimensional Kalman filtering technique,we can establish a recursive framework for designing the learning gain matrix along both time and iteration axes and prove the input error converges to zero asymptotically in mean square sense.Furthermore,we consider the case that prior distribution of nonuniform trial lengths is unknown,a suboptimal algorithm is given by an asymptotic estimation method.