提出了一种机械手的神经网络直接离散时间自适应控制算法。该算法是神经网络方法和自适应动态滑动模控制方法的集成。自适应动态滑动模控制的作用有两个：其一是在神经网络控制失灵的情形下提供控制系统的全局稳定性；其二是改善系统的跟随性能。整个系统的全局稳定性和跟随误差的收敛性采用李雅普诺夫稳定性理论进行了证明，并得到了一种新颖的神经网络权值调整算法。 A neural network direct discrete-time adaptive control algorithm for manipulator is proposed. This algorithm is an integration of neural network method and adaptive dynamic sliding mode control method. There are two roles of adaptive dynamic sliding mode control: one is to provide the global stability of the control system under the control of the neural network; and the other is to improve the system following performance. The global stability of the whole system and the convergence of following error are proved by the Lyapunov stability theory, and a novel algorithm for weight adjustment of neural networks is obtained.