基于神经网络的自适应非线性控制及仿真研究

研究了神经网络非线性动态系统的自适应控制方法,首先利用改进的非线性自回归滑动平均模型,采用多层前向神经网络辨识非线性系统模型,然后直接由辨识结果设计出控制器,并根据控制误差对控制律作在线修正.利用导弹模型进行了控制仿真,仿真结果表明采用此方法可以得到较好的控制效果,而且在模型不确定和有噪声干扰的情况下仍能正常跟踪给定的迎角信号,具有较好的鲁棒性.

Artificial-neural-network-based nonlinear adaptive control and simulation

The increasing complexity of the nonlinear systems highlights the need for advanced modeling and control techniques for effective control. An artificial-neural-network-based nonlinear adaptive control system was studied, which identified the dynamics of the plant with improved NARMA(nonlinear auto-regressive moving average) model at first. Then, the control algorithm was deduced with respect to the model directly. An adaptive system was designed to control a longitudinal missile autopilot which achieves tracking of external reference commands in angle of attack.Simulation results illustrate the ability of this technique to compensate for unmodeled dynamics and uncertain plant nonlinearities.