基于神经网络动态逆的微小型飞行器半实物仿真

Hardware-in-the-loop Simulation of MAV with Dynamic Inversion of Neural Network

针对微小型飞行器易受外界干扰,难于控制的特点,在经典PID控制的基础上,引入神经网络动态逆方法,设计复合控制律。基于Matlab建立微小型飞行器的Simulink仿真模型。控制器、舵偏角传感器等物理实物通过RS232串口连接,搭建半实物仿真环境。仿真结果表明,该方案能有效地控制微小型飞行器按照预定航线,以稳定的姿态飞行,具有较强的鲁棒性。

The micro-aerial vehicle（MAV） is easy to be disturbed by exterior factors and hard to control. Based on traditional PID control methods, a multiple control scheme is designed by importing dynamic inversion of neural network. A Simulink model of MAV is built by Matlab and hardware-in-the-loop Simulation is constructed through RS232, by controller, sensor of rudder angle and other physical hardware. The simulation outcomes demonstrate that the scheme not only can effectively control the MAV with a steady attitude flying in a desired course, but also has a strong robustness.