微粒群优化倒立摆神经网络控制器设计研究

Research and Design of Inverted Pendulum Neural Network Controller Based on Particle Swarm Optimization Algorithm

将微粒群算法和多层前馈神经网络相结合,提出了一种利用微粒群算法代替BP算法训练多层前馈神经网络权值,以实现神经网络控制的方法,并对非线性模型的辨识问题和一级直线倒立摆的控制问题进行了仿真研究.仿真实验表明:微粒群算法在神经网络控制及非线性模型辨识方面效果良好,具有良好的应用前景.

This paper describes a neural network control approach for the use of particle swarm optimization （PSO） instead of using BP algorithm to train the networks to construct a PSO neural network. The combination of PSO with neural network is adopted to train multilayer feedforward neural network to identify a nonlinear model and an inverted pendulum controlled. Simulation results show that the proposed algorithm has good dynamic and static control performance for inverted pendulum and have excellent performance on nonlinear model identification. The proposed approach is a kind of algorithm that can be widely used in neural network control and nonlinear model identification.