Research On Adaptive Neural Network Quantized Control Algorithm For Nonlinear Models

Due to the theoretical and practical significance of quantized control algorithms in the research of digital control systems and networked control systems,the algorithms have received a lot of attention in recent years.Sensor faults and input delay exist extensively in real nonlinear system models and seriously affect the control performance of the models.Therefore,the research on stabilization for nonlinear models with quantized information and sensor faults is crucial.With the extreme growth of computer control method and the gradual maturation of control theory,the quantization control problem for nonlinear system models has attracted numerous scholars’ attention.However,the majority of current results only consider input quantization of models,and there are few studies on fault-tolerant control of nonlinear models with states quantization.To address this problem,this paper studies the adaptive regulation problems for two classes of nonlinear models with quantized states and sensor faults,the main work is as follows:(1)For a class of nonlinear models with quantized states and sensor failure,the adaptive stabilization regulation problem is investigated.The neural network is introduced to handle continuous unknown functions in the system model online.By improving the traditional adaptive backstepping design algorithm to compensate for the effects of quantized states,a new controller is constructed using quantized states under sensor faults.Under this controller,the boundedness of all signals in this nonlinear model can be guaranteed.At last,the effectiveness of the presented algorithm is demonstrated by a Matlab simulation experiment.(2)For a class of input-delayed stochastic models with sensor failure and simultaneous quantization of states and input,the adaptive fault-tolerant control problem is studied.First of all,a traditional neuroadaptive fault-tolerant control algorithm is developed to tackle the stabilization problem in this stochastic model with unquantized information.Secondly,two new fault-tolerant control algorithms are proposed to solve the discontinuity problem in this stochastic model with only states quantization and both input and states quantization,respectively,so as to ensure the stability of this stochastic model.Meanwhile,the neural network approximation approach and the Pade approximation technique are introduced to deal with unknown terms and input delay in this system model,respectively.Under the presented controller and quantizer,all closed-loop signals in this stochastic model maintain bounded by probability.Finally,the feasibility of the presented algorithm is demonstrated by a Matlab simulation experiment.