Rotor craft attitude control method based on genetic algorithm optimization neural network

Abstract

The invention relates to a rotor craft attitude control method based on a genetic algorithm optimization neural network. The rotor craft attitude control method comprises: an initial population T is established and data are preprocessed; a coding mode is determined; a fitness function is determined; the genetic operation of a genetic algorithm is performed; an obtained optimal solution is substituted into a weight and a threshold of a BP neural network model; an error is calculated, and if the requirement is not met, the weight and threshold of the neural network model are updated by using anerror back propagation algorithm; if the requirement is met, the output of the BP neural network model is applied to a PID controller and the PID controller adjusts an inertia link coefficient Kp, anintegral link coefficient Ki and a differentiation link coefficient Kd to adjust the control intensity and effect and outputs an ideal control signal; and the attitude of the rotorcraft is ideally controlled and thus the ideal flying attitude is restored. Therefore, a defect that the neural network falls into a local minimum value is overcome effectively.

Description

technical field [0001] The invention relates to the technical field of rotorcraft attitude control, in particular to a method for controlling the attitude of a rotorcraft based on genetic algorithm optimization neural network. Background technique [0002] Traditional PID controllers have been widely used because of their simple principle, mature technology, strong robustness, and suitability for industrial production sites with harsh environments. However, traditional PID control is based on accurate models, and system characteristics change and control quantities There is a linear mapping relationship between them. If a conventional PID controller is used, a set of fixed parameters is used to adapt to some parameter changes and interfere with many control systems, it will not be able to obtain satisfactory control effects, and even when the parameter variation range is too large, the system performance will be obvious. worse. Therefore, the traditional PID control cannot...

AI Technical Summary

Problems solved by technology

[0004] However, in the process of using the neural network, it is found that the neural network is easy to fall into the local minimum value in the process of finding the optimal solution because of the gradient descent method.

Although the BP network, especially the application of various second-order gradient algorithms, uses weight search to speed up the convergence, but when the objective function has many local minima in the weight space, gradient search still cannot effectively avoid falling into local minima. The problem

Images