基于对抗网络的六旋翼无人机串级线性自抗扰控制系统设计

Design of cascade linear active disturbance rejection control system for UAV based on machine learning

摘要:

当无人机行进轨迹内存在明显转向行为时，若不能实现姿态角与响应曲线的有效耦合，则会使飞行器的抗扰能力下降，从而降低无人机飞行品质。为解决上述问题，设计基于机器学习的无人机串级线性自抗扰控制系统。按需连接串级线性跟踪微分器、扩张状态观测器、自抗扰型无人机姿态控制器与行进位置控制器，完成无人机串级线性自抗扰控制硬件系统的设计。建立机器学习模型对抗网络，求解无人机串级线性动力学运动公式，联合相关运动数据，完成基于机器学习的无人机串级线性动力学性能分析。定义串级线性位姿坐标，通过推导自抗扰运动节点矩阵的方式，计算具体的自抗扰性控制条件，实现对无人机串级线性位姿的自抗扰性控制，联合相关应用部件结构，完成基于机器学习的无人机串级线性自抗扰控制系统的设计。实验结果表明，机器学习型控制系统作用下，俯仰角、滚转角两类姿态角与标准响应曲线之间的耦合误差均不超过10%，即便在行进轨迹内存在明显转向行为的情况下，应用该系统也可以实现姿态角与响应曲线的有效耦合，能够有效保障无人机飞行器的抗扰能力。

Abstract:

When there is obvious turning behavior in the trajectory of the drone, if the effective coupling between attitude angle and response curve cannot be achieved, the aircraft's anti-interference ability will decrease, thereby reducing the flight quality of the drone. To address the above issues, a machine learning based cascade linear autodisturbance rejection control system for unmanned aerial vehicles is designed. Connect the cascade linear tracking differentiator, extended state observer, auto-disturbance rejection unmanned aerial vehicle attitude controller, and travel position controller as needed to complete the design of the drone cascade linear auto-disturbance rejection control hardware system. Establish a machine learning model adversarial network, solve the drone cascade linear dynamic motion formula, combine relevant motion data, and complete the drone cascade linear dynamic performance analysis based on machine learning. Define the cascade linear pose coordinates, calculate the specific auto-disturbance rejection control conditions by deriving the auto-disturbance rejection motion node matrix, and achieve auto-disturbance rejection control of the drone cascade linear pose. Combined with relevant application component structures, complete the design of the drone cascade linear auto-disturbance rejection control system based on machine learning. The experimental results show that under the action of a machine learning control system, the coupling error between the pitch angle and roll angle of the two types of attitude angles and the standard response curve does not exceed 10%. Even in the case of obvious turning behavior in the travel trajectory, the application of this system can achieve effective coupling between the attitude angle and response curve, effectively ensuring the anti-interference ability of the unmanned aerial vehicle.