加入预测信息的反馈误差学习模型及其仿真研究

针对非线性平衡控制问题，提出了一种加入预测信息的反馈误差学习（P-FEL）模型，该模型使用系统状态预测信息和反馈控制器的输出信号共同构成前馈神经网络控制器的教师信号，使用在线BP算法保证运动控制和运动学习同步进行。将P-FEL模型应用于倒立摆平衡控制的仿真实验结果表明，P-FEL模型可以有效地减少前馈神经网络控制器对反馈控制器参数的依赖性，同时还具有良好的平衡控制性能和鲁棒性。

Design and Simulation of Predictive Feedback Error Learning Model

Aiming to cope with the nonlinear balance control problem, the predictive Feedback Error Learning (P-FEL) model was proposed. In the P-FEL model, both the state predictive signal and the output of the Conventional Feedback Controller (CFC) were integrated into the teacher signal of the Neural Network Feedforward Controller (NNFC). An on-line back-propagation (BP) algorithm with the self-adaptive learning rate was developed and employed in the NNFC to realize the combination of learning and controlling. Computer simulations on inverted pendulum balancing task demonstrate that the P-FEL model could effectively reduce the precision requirements of the CFC parameters, and guarantees good balance performance and acceptable robust performance.