基于分段线性逼近网络的直接自适应控制

对一类单输入单输出系统的控制问题进行了研究 ,设计了对这类系统的直接自适应控制方案。控制方案的一部分是采用分段线性逼近神经网络逼近给定的非线性系统 ,利用已经发生的实际数据对网络的权值进行训练 ;另一部分是结合神经网络的一步预测输出和系统的参考输出 ,通过梯度优化器计算得到系统的控制输入。数值仿真结果表明 ,控制效果比较理想

Direct Adaptive Control Based on the Networks of Piecewise Linear Approximators

The control of a class of SISO systems is studied in this paper.The direct adaptive control scheme for this class of SISO systems is designed.This scheme first approaches the known nonlinear system by the networks of piecewise linear approximators and trains the weighted values of networks by using known data; then computes the system control input through gradient optimizer based on the system reference output and one-step forecast output of neural networks.Numerical simulation shows that the control effect is perfect.