基于深度神经网络的直线二级倒立摆控制器设计

Controller Design of Linear Two-stage Inverted Pendulum Based on Deep Neural Network

针对直线二级倒立摆抗干扰控制器设计问题,研究了基于深度神经网络的智能控制方法。首先介绍了BP神经网络和深度神经网络模型及优化算法,并且根据直线二级倒立摆状态方程,研究了基于深度神经网络的直线二级倒立摆控制算法。然后设计了一个六输入单输出的深度神经网络控制器模型,并利用Pytorch框架,以LQR作为导师进行神经网络的训练,训练完成后利用MATLAB软件对训练后的神经网络进行仿真实验验证,并与BP神经网络控制器进行对比,最后在直线二级倒立摆实验平台上进行实验验证。仿真与实验表明,所设计的深度神经网络控制器能够实现直线二级倒立摆的良好抗干扰控制,从而证明了该研究设计方法的合理性和有效性。

Aiming at the design of anti-jamming controller for linear two-stage inverted pendulum,an intelligent control method based on deep neural network is studied.In this paper,BP neural network and deep neural network models and optimization algorithms are introduced.According to the state equation of linear two-level inverted pendulum,the control algorithm of linear two-level inverted pendulum based on deep neural network is studied.By designing a deep neural network controller model with six inputs and one output,the Pytorch framework is used to train the neural network with LQR as a mentor.The network controller is compared,and finally the experimental verification is carried out on the linear two-stage inverted pendulum experimental platform.Simulation and experiments show that the designed deep neural network controller can achieve good anti-interference control of the linear two-stage inverted pendulum,which proves the rationality and effectiveness of the design method in this paper.