Adaptive Neural Network Control For Uncertain Nonlinear Systems With Input Constraints

In real world,most controlled systems are intrinsically nonlinear,uncertain,and often affected by input constraints.Hence,the study of uncertain nonlinear systems with input constraints has received much attention.However,there is no universal method or theory for these systems,it is necessary to study the input constraints and control problems by developing particular methods.Based on the existing works,this thesis utilizes the neural networks' approximation theory to investigate two control problems with input constraints.One is the problem of adaptive control for stochastic nonlinear systems;the other one is the distributed coordinated control problem of nonlinear multi-agent systems.The main contents are summarized as follows:1.The adaptive neural network stabilization control problems are investigated for uncer-tain stochastic nonlinear systems with time-delay and input constraints,where different input constraints are considered.For input saturation,we propose the adaptive neural network control scheme for single input single output stochastic nonlinear time-delay systems.For dead-zone input,the adaptive neural network decentralized controller for stochastic nonlinear large-scale systems with time-delay is developed.The proposed control schemes guarantee that all signals in the closed-loop systems are bounded in probability.Several examples are provided to demonstrate the applicability of the pro-posed design schemes.2.The problems of adaptive neural network tracking control for uncertain switched stochas-tic nonlinear systems with input constraints are investigated.First,we consider the strict-feedback switched stochastic nonlinear systems with input saturation.By combining the neural networks' approximation theory,backstepping design method and common Lyapunov function method,an adaptive neural network tracking control scheme is de-veloped.The proposed control scheme guarantees that all signals in the closed-loop sys-tems are bounded in probability,and the tracking error converges to a small neighbour-hood of origin.Second,an adaptive neural network tracking control scheme is proposed for pure-feedback switched stochastic nonlinear systems with unknown backlash-like hysteresis,where the mean-value theorem is used to transform the non-affine functions to affine form.Numerical examples are provided to demonstrate the effectiveness of the proposed design methods.3.The distributed consensus problem for nonlinear multi-agent systems with input satu-ration is investigated.Neural networks are utilized to identify the unknown nonlinear functions,and an auxiliary system is introduced into the control design to compen-sate the effect of input saturation.By incorporating the command filtered technique into the backstepping design framework,a distributed consensus control scheme is con-structed recursively.Using the Lyapunov stability theory,it is proved that all signals in the closed-loop systems are semi-globally uniformly ultimately bounded and the con-sensus errors converge to a small neighbourhood of origin.Furthermore,the problem of distributed containment control for a class of nonlinear multi-agent systems with actua-tor failures is considered,and effective design methods for desired distributed controller are developed.Simulation results show the effectiveness of the proposed control ap-proaches.4.The problem of prescribed performance distributed consensus control for non-affine nonlinear multi-agent systems with unknown dead-zone input is investigated.Neural networks are utilized to approximate the unknown nonlinearities.By introducing pre-scribed performance,the transient and steady performance of synchronization errors are guaranteed.Based on Lyapunov stability theory and the dynamic surface control technique,a distributed consensus control scheme without depending on the dead-zone parameters for nonlinear multi-agent systems is proposed,which ensures semi-globally uniformly ultimately boundedness of all signals in the closed-loop systems and enables the output of each follower to synchronize with the leader within predefined bounded tracking error.A numerical example is provided to demonstrate the effectiveness of the proposed control scheme.