基于贝叶斯改进神经网络的电力无人机鲁棒姿态控制方法

针对电力无人机在工作状态下受到外部因素干扰、无法精准控制运动姿态的问题,提出基于贝叶斯改进神经网络的电力无人机鲁棒姿态控制方法。综合考虑电力无人机的组成结构、运动以及动力原理,构建电力无人机数学模型。利用传感器设备检测电力无人机的实时位姿,采用飞行路线规划的方式确定姿态控制目标。在考虑风场威胁条件和故障状态的情况下,利用贝叶斯改进神经网络计算无人机的姿态控制量,以鲁棒姿态控制器作为硬件支持,实现鲁棒姿态控制。通过性能测试得出结论:优化设计方法的姿态角控制误差始终低于0.2°,且在三种不同风场工况下,控制误差的波动程度不高于0.5°,与传统方法相比,优化设计方法在姿态控制精度和鲁棒性方面具有明显优势。

Robust attitude control method of electric UAV based on Bayesian improved neural network

Aiming at the problem that the electric UAV cannot precisely control its motion attitude due to the interference of external factors, a robust attitude control method based on Bayesian improved neural network was proposed for electric UAV. The mathematical model of electric UAV is established by considering the composition, motion and dynamic principle of electric UAV. The sensor equipment is used to detect the real-time pose of the electric UAV and the attitude control target is determined by the way of flight path planning. Considering the wind threat condition and fault state, the attitude control quantity of UAV is calculated by using Bayesian improved neural network, and the robust attitude controller is used as the hardware support to realize the robust attitude control. Through the performance test, it is concluded that the attitude Angle control error of the optimized design method is always lower than 0.2°, and the fluctuation degree of the control error is no more than 0.5° under three different wind conditions. Compared with the traditional method, the optimized design method has obvious advantages in the accuracy and robustness of attitude control.