Research And Application Of Variable-Parameter Finite-Time Recurrent Neural Network Based On Neurodynamic

In practical engineering and scientific computation,many problems can be studied as a class of systems,traditional numerical methods can not meet the requirements of real-time,online,fast,and accurate for solving problems because of the serial processing ability,so the neural network with high-performance parallel processing ability has gradually become a powerful tool to solve this kind of problem.In addition,the existence of periodic noise will affect the performance of the model in practical application and may cause calculation deviation or even result in calculation failure.Therefore,both the time-varying dynamical system solution with periodic noise and its application in practical engineerings,such as redundant manipulator tracking tasks,are worth exploring.Based on the recurrent neural network of neurodynamics,this paper applies it to solving time-varying linear matrix equations,time-varying convex quadratic programming problems,and repetitive motion planning of redundant manipulators,the specific research contents are as follows:1.A variable-parameter finite-time neural network based on neural dynamics is proposed to solve time-varying linear matrix equations.Based on the variable parameter recurrent neural network model,the finite-time convergence problem is considered,and the convergence and robustness of the method are analyzed theoretically,it is verified that the variable-parameter finite-time neural network has better convergence and robustness.2.In order to solve the disturbance of periodic noise,a variable-parameter finitetime circadian rhythms neural network model is proposed,and it is used to solve the time-varying convex quadratic programming problem with periodic noise.The stable convergence performance under noise is proved theoretically,and the simulation results are compared with the variable-parameter finite-time neural network and two other known models,the experimental results show that the variable-parameter finite-time circadian rhythms neural network is stronger robustness to periodic noise.3.A variable-parameter finite-time circadian rhythms neural network is applied to repetitive motion planning of redundant manipulators under periodic noise.Firstly,the tracking problem of redundant manipulators is transformed into a convex quadratic programming problem,then the convex quadratic programming problem is transformed into a matrix equation,and finally,the matrix equation of the tracking problem is solved by using the proposed model.The proposed variable-parameter finite-time circadian rhythms neural network and variable-parameter finite-time neural network are simulated and compared,it is verified that the variable-parameter finitetime circadian rhythms neural network has better convergence and robustness in solving end-effector tracking tasks.The validity,robustness,and practicability of the proposed variable-parameter finite-time circadian rhythms neural network are further verified by two physical experiments.