一种基于动态逆的控制方案在无人机中的应用研究

动态逆技术通过成功的实际飞行测试而获得广泛认同，目前该研究热点在于对象模型的不确定性的抑制及鲁棒性的研究。文中提出一种基于在线神经网络的新型动态逆控制器，其主要思想是通过在线迭代的神经网络最大程度地挖无人机非线性不确定模型及其逆模型，同时通过一个梯度修正环节来补偿误差的影响。仿真验证包括三部分：对极快变量回路的仿真、航迹角跟随的仿真以及三维空间中对给定轨迹的跟踪。仿真结果表明，该方案能够解决精确的

Dynamic Inversion Based Control Scheme Applied to Pilotless Aircraft

While the field involving pilotless aircraft is increasingly widening, challenge is also existing in the flight control system. The dynamic inversion technique is approved by kinds of flight tests recently. Now the research focus of this method lies in the compensation of the effect on the model uncertainty and the robustness. A new dynamic inversion controller based on online neural networks is presented and verified by a precise 3-D path following problem. The key idea of the method is that a modified unit by gradient method is used to compensate the error while using online neural networks to form dynamic inversion controller. The advantage of the method is that the converzence can be achieved even the error is large. This is that because both the online neural network inversion controller and the gradient modified unit may elimiate the error at the same time.