On Adaptive Iterative Learning Control And Kalman Consensus Filtering With Applications In High-speed Train Operation Control

This thesis focuses on adaptive iterative learning control and Kalman consensus filtering with their applications in high-speed train operation control.Some practical constraints are addressed from speed tracking control and estimation perspectives,respectively.The main contents and contributions are summarized as follows:1.Adaptive iterative learning control(AILC)strategies for high-speed trains with unknown speed delays and control input saturation are designed to address the speed trajectory perfect tracking problem.The train motion dynamics containing nonlinearities and parametric uncertainties are formulated as a nonlinearly parameterized system.First,by using the parameter separation technique,a partially saturated adaptive iterative learning controller is constructed for input saturation and a fully saturated parameter updating law is presented for iterative parameter identification.Consider unknown time-varying speed delays as well,a delay compensate term is integrated into the AILC to alleviate the effect of delays.Based on Lyapunov functional analysis,it is rigorously proved that the proposed AILC mechanism can guarantee L_[0,T]~2 convergence of train speed to the desired profile during operations repeatedly.Cases studies with numerical simulations further verify the effectiveness of the proposed approach.2.Adaptive iterative learning reliable control(AILRC) schemes are developed in this thesis for high-speed trains subject to unknown time-varying state delays and input saturation as well as traction/braking faults.In regard to nonlinearly parameterized uncertainties of train model,not only the nonlinearly parameterized controlled object but also the nonlinearly parameterized input distribution matrix is investigated.Without the need for precise system parameters or analytically estimating bound on actuator faults variables,the novel data-driven AILRC is constructed by a nonlinear feedback term and a robust term.The nonlinear influence brought by actuator faults,input saturation and state delays can be compensated with the resultant algorithms.The proposed AILRC can be applied not only to SISO high-speed train systems,but also to a class of MIMO nonlinearly parameterized systems.It is shown that the L_[0,T]~2 convergence of SISO and MIMO systems is proved through a new time-weighted Lyapunov-Krasovskii-Like composite energy function(CEF).The validity of the proposed AILRC is further verified by simulation.3.Poor local sensor node estimates caused by limited observability,network topologies that restrict allowable communications,and communication noises between sensors are challenging issues not yet fully resolved in the framework of distributed Kalman consensus filters.This thesis confronts these issues by introducing a novel distributed continuous-time information-weighted Kalman consensus filter(CT-IKCF)algorithm for state estimation of dynamic linear systems in sensor networks.A new measurement model is selected that only depends on local information available at each node based on the prescribed communication topology,wherein all the estimates of neighbor sensors are weighted by their inverse-covariance matrices.Locally optimal solutions are then derived for the proposed distributed CT-IKCF considering channel noises in the consensus terms.It is formally proven by Lyapunov techniques that,using the new distributed CT-IKCF,the estimates of all sensors reach converge to consensus values that give locally optimal estimates of the target.Moreover,if the target has a nonzero control input,a method is giving of incorporating estimates of the target's unknown input.Simulation case studies show that the proposed distributed CT-IKCF works well for limited observability and outperforms other methods in the literature.4.A class of distributed discrete-time Kalman consensus filtering techniques are proposed to deal with dynamic target's state estimation of linear or nonlinear systems.For discrete-time linear time-varying systems in sensor networks,a distributed information-weighted Kalman consensus filter(DT-IKCF)is firstly constructed based on novel measurement models using the topology of information flow.Moreover,for nonlinear high-speed train dynamics,an distributed data-driven discrete-time information-weighted Kalman consensus filter(DD-IKCF)for speed estimation is investigated to overcome technical difficulties in optimal state estimation of distributed KCF for nonlinear models.By means of full form dynamic linearization(FFDL),a class of nonlinear model can be transformed into a TARX model with unknown time-varying parameters equivalently.Based on iterative Kalman filter(IKF)for system identification,the distributed DD-IKCF can be applied to high-speed trains for synchronization of estimates from all the sensors to a common actual speed in a least-square sense,which is further verified by simulation results.