不确定性X-Y定位平台自学习控制研究

Self-Learning Control for X-Y Position Table with Uncertainty

针对带不确定性的X-Y定位平台系统位置控制问题,提出了径向基函数神经网络的自学习控制策略。首先建立X-Y定位平台轴系统的动力学模型,然后利用增广变量法设计了基于神经网络PID控制器,利用RBF神经网络良好的逼近能力来进行自学习控制,设计了改进随机梯度算法来实现网络权值的自适应调整,并加快其学习速度。针对神经网络动态特性欠缺的问题,设计了PID控制器来保证控制阶段初期的跟踪精度。最后通过仿真详细分析了其控制机理,并证明了该方案的有效性,具有较高工程应用价值。

An adaptive radical basis function neural network （RBFNN） control scheme based on learning on line for X-Y position table with uncertainty was put forward here.Firstly,X-Y position table system dynamics model was established,and the augmented variable method was designed based on the neural network PID controller,RBF neural network good approximation ability was used to get self learning control Improved stochastic gradient algorithm was proposed to realize the network weights adaptive adjustment,and speed up the learning speed.PID controller was designed to ensure control of the early stage of tracking precision.According to the neural network dynamic characteristic short question,PID controller was used as assistant direction error controller.Finally,Control mechanisms were analyzed in detail by the simulation results.The simulation resulted proves the effectiveness of the scheme,which has high value of engineering application.