针对一般非线性不确定系统设计了一种e-修正神经网络直接自适应控制方法。首先采用虚拟控制量的方法, 并将其分解成参考模型输出、线性动态补偿输出与神经网络自适应输出三项；然后针对传统σ-修正神经网络在权值更新时的不足, 设计了一种基于e-修正方法的权值自适应更新律, 并设计了输出反馈误差观测器用以对神经网络进行训练；最后对基于σ-修正与e-修正两种权值自适应更新律进行仿真对比。仿真结果表明基于e-修正神经网络方法在跟踪误差、不确定性逼近等效果上均优于基于σ-修正神经网络方法。

A direct adaptive control method was proposed based on e-Modification Neural Network(NN) for a general nonlinear uncertain sysytem.A method called pseudo-control was introduced which was constructed from the outputs of reference models the linear dynamic compensator outputs and NN outputs.To solve the proplem of the σ-Modification NN in weight updating we proposed the e-modification method to design the NN update law and designed an error observer based on output feedback to train the NN.Simulation was made for direct adaptive control of both e-Modification NN and σ-Modification NN.The simulation results show that:compared with the mthod based on σ-Modification NN the designed controller based on e-Modification NN has a better performance on output tracking error and uncertainty approaching.