基于神经网络的无人机负载模拟器的复合控制

衡量负载模拟器系统性能的关键指标是多余力矩的抑制。针对无人机负载模拟器系统的非线性及多余力矩强扰动的特点,依据神经网络的非线性逼近和自学习特性,提出了一种基于神经网络和前馈相结合的复合控制器,用来提高系统的性能。复合控制器利用前馈来补偿定常多余力,利用神经网络进行在线辨识、控制来补偿系统的非线性部分,很好地抑制了多余力矩。该文给出了具体的控制结构和算法。仿真结果还表明该方法极大地改善了系统动态加载性能,有很强的鲁棒性。

Load Simulator for Unmanned Aerial Vehicle Hybrid Control Based on Neural Network

How to eliminate the surplus torque of a loading system is one of the key problems to design a load simulator. Due to ANN's non-linear approximation and self-learning characteristics, a hybrid controller based on adaptive neural network and feedforward is proposed for the system of load simulator for unmanned aerial vehicle. In the controller, feedforward is used for the compensation of constant linear disturbances, and neural network is adopted for online identification and the compensation of the nonlinear disturbances, which eliminates the surplus torque effectively. The structure and algorithm of controller are presented here. The simulation results show that the load simulator's dynamic performance is greatly improved, and the robustness to unknown external load disturbances is improved.