Maxout模糊神经网络四旋翼无人机复合控制器

Research on Fuzzy Neural Network Four-Rotor UAV Compound Controller Based on Maxout Identification

针对常规PID控制下四旋翼无人机抗干扰性弱的问题,提出了一种集合了常规PID和模糊理论以及神经网络算法的复合控制器。通过将误差信号模糊化输入神经网络进行自适应控制,将角度和角速度误差作为收敛条件,提高了无人机的抗干扰能力以及响应的精确性。同时搭建基于Maxout激活函数的神经网络模型辨识器,实现了对被控模型的精准识别。仿真结果表明,结合模糊神经网络的复合控制器相对于常规PID控制器,无人机抗干扰能力大幅度提升,新的控制方式具有更强的抗干扰性和鲁棒性。

Aiming at the problems of weak anti-interference of four-rotor UAV under conventional PID control, acomposite controller combining conventional PID,fuzzy theory and neural network algorithm is proposed. By fuzzifyingthe error signal into the neural network to adaptive control and using the angle and angular velocity errors as the con-vergence condition, the anti-jamming capability and the accuracy of the response of the UAV was improved. At thesame time, a neural network model recognizer based on the Maxout activation function was built to achieve accuraterecognition of the controlled model. Simulation experiments show that the composite controller combined with fuzzyneural network can greatly improve the anti-interference ability of the drone compared with the conventional PID con-troller, and the new control method has stronger anti-interference and robustness.