Neural network backstepping control method based on error training

Abstract

The invention discloses a neural network backstepping control method based on error training, solves the problems that an existing neural network backstepping control method is low in convergence speed and a neural network cannot accurately estimate unmodeled dynamics, and consequently system tracking errors are large, and belongs to the field of neural network backstepping control methods of nonlinear systems. The method comprises the steps that S1, a non-linear n-order system state space model containing unmodeled dynamics is established, and the state variable is [x1,..., xn] T; s2, error variables z1 and zi are determined, z1 = x1-yd, zi = xi-alphai-1, and alphai-1 represents a virtual control function; s3, a differential estimator of the error zi is established, the input of the differential estimator is zi, and the output of the differential estimator is the estimation of zi; s4, calculating the estimation error of the current radial basis function neural network obtained in S3, and carrying out gradient descent training on the weight of the neural network based on the estimation error, and S5, calculating a control input signal of the nonlinear system according to alpha n.

Description

technical field [0001] The invention belongs to the field of neural network backstep control method of nonlinear system. Background technique [0002] Neural network backstepping control is a method in nonlinear system control. Its basic principle is to use the characteristics that neural network can approach any unknown function with a certain error to estimate the unmodeled dynamics in the system. In the design process of backstepping method The estimated value of the neural network is fed back to the nonlinear system to reduce the disturbance of the system caused by the unmodeled dynamics. At present, most of the neural network backstep control adopts the final consistent boundedness and the neural network weight update strategy based on σ adjustment. Ultimately consistent boundedness is a kind of stability defined in infinite time. The convergence speed of the system designed by this method is generally slow, and it is not suitable for some control systems with stricter...

AI Technical Summary

Problems solved by technology

[0003] Aiming at the problems that the existing neural network backstepping control method has a slow convergence speed and the neural network cannot accurately estimate the unmodeled dynamics, which leads to large system tracking errors, the present invention provides a neural network backstepping control method based on error training

Images