基于神经网络的无人直升机姿态控制系统设计

 首先根据模型参考自适应控制理论,将模型逆与在线神经网络结合,设计了神经网络自适应姿态控制系统。接着叙述反馈线性化及模型逆理论,分析系统的模型跟踪误差动力特性,设计神经网络控制器及在线算法。然后以某无人直升机俯仰通道为例,对神经网络姿态控制系统进行仿真。结果表明该系统能够对未建模特性、参数不确定性等引起的模型逆误差进行自适应,而且在传感器输出中具有白噪声时仍然能够获得较好的响应特性。

Design of Attitude Control System Based on Neural Network to Unmanned Helicopter

The neural network adaptive attitude control system is designed based on model reference adaptive control theory and combination of online neural network with augmented model inversion.Then feedback linearization and model inversion theories are described,model tracking error dynamics is analyzed and the neural network controller and the online learn laws are designed.Finally,the attitude control system of pitch channel with neural network is designed and simulated as an example.The results show that this neural network control system adapts to model inversion error originated from model and parameter uncertainties,and that the responses are still satisfactory in case of the presence of sensor output white noise.