Abstract: The small unmanned helicopter is a highly nonlinear system with significantly dynamic coupling and uncertainty parameters. This paper presents an adaptive back-stepping control method based on CMAC（Cerebellar Model Articulation Control）. The proposed method uses CMAC to learn the system uncertainty online and derivative information of virtual controllers in each stage of back-stepping control. In order to reduce the system uncertainty of CMAC, a robust controller is presented and the control law is derived via recursive regression of back-stepping method. The simulation results demonstrate that, with the parameter uncertainty and error of the model, the proposed control law has the excellent attitude tracking performance and robustness.

Key words: small-scale unmanned helicopter, neural network, back-stepping adaptation control, robust

摘要： 针对具有强非线性、高度耦合以及参数不确定性特点的小型无人直升机系统，提出一种基于小脑模型关节控制器（Cerebellar Model Articulation Control，CMAC）神经网络的自适应反步控制方法，该方法采用小脑模型关节控制器神经网络在线学习系统不确定性以及反步控制中各阶虚拟控制量的导数信息，设计鲁棒控制项克服CMAC神经网络在线学习系统不确定性的误差，控制律由反步法回归递推得到。仿真结果表明，在模型参数不确定和存在较大误差的情况下，所设计的控制律具有理想的姿态跟踪性能以及良好的鲁棒性。

关键词: 小型无人直升机, 神经网络, 反步自适应控制, 鲁棒