Abstract: Considering trajectory tracking of a class of free-floating space robot with disturbance and model uncertainties, a neural network adaptive control strategy which based on the global model approximation is proposed for space robot. The control strategy uses neural networks to approach the uncertain model under uncertain gravity environment by global approximation, adjust learning the uncertain part of system on-line. The neural network approach errors and outside bounded disturbance can be eliminated by a robust controller. The control strategy can not depend on the model, simplifies the structure of control system, and can be used under uncertain gravity environment by the same controller. The control strategy can guarantee the stability of system and the asymptotic convergence of tracking errors based on the Lyapunov theory. The simulation result that this control strategy under uncertain gravity environment is effective.

Key words: neural network, uncertainty, space robot, global approximation, adaptive control

摘要： 针对具有模型不确定性以及外部干扰下的自由漂浮空间机器人，采用一种整体逼近的神经网络自适应控制方法。该方法采用RBF神经网络对不同重力环境下系统模型的不确定项进行整体逼近，对系统的不确定项进行在线自适应学习。神经网络的逼近误差以及外界干扰由鲁棒项进行消除。该方法不依赖于系统模型，简化了控制系统的结构，在考虑重力等不确定项的情况下不用改变控制器也能进行控制，并且根据李亚普诺夫理论证明了所设计控制器使系统渐进稳定。在不同重力环境下进行了仿真，验证了控制方案的有效性。

关键词: 神经网络, 不确定性, 空间机器人, 整体逼近, 自适应控制