中文摘要：

自由飘浮空间机器人是一种复杂的非线性系统,当考虑各种因素引起的模型不确定性和外部干扰时,传统控制已不能满足要求。为此,文章考虑不确定性干扰对系统的影响,基于神经网络在线建模技术,将系统误差和误差变化率作为状态量,利用李雅普诺夫方程设计网络权值变化率,通过训练权值对不确定部分进行逼近,在此基础上设计神经网络自适应控制项,用于修正基于名义模型的控制律,补偿不确定项造成的控制误差,并证明了控制系统的稳定性。7关节空间机器人的数值仿真结果表明,文中所设计的自适应神经网络控制器,在机械臂末端有干扰和负载的情况下,也能向期望轨迹靠拢,并在一定误差范围内保持收敛,相比于传统PID控制器,性能更优。

英文摘要：

Free-floating space robot system is a very complicated nonlinear analysis. Traditional control strategy cannot meet the technical requirements when considering uncertain terms and external disturbances. For this reason, the influence of the uncertainty to the system is considered in this article. This article regard the control er- ror and the error rate as system state. Basing on neural network on-line modeling technique, design the change rate of network weights and approach the uncertain parts by weight of training. Then design adaptive neural network control item, which can be used to amend nominal-model-based control and compensate the control error caused by uncertain terms. Meanwhile, The stability of control system is proved. The numerical simulation of 7 DOF （ degree of freedom） space robot show that with the adaptive neural network controller proposed in this article, the end-effector with disturbances and load can also close to the desired trajectory and converge when the error within a finite range. The controller performs better than traditional PID controller.