说明：  摘要：给出了解决机器人控制问题一种神经网络方法。使用一个分级神经网络(NN)结构学习刚体机器人动力学特点。对于一般类别的机械手，使用前训练一系列的三层前馈网络模块，然后把这些基函数实时地用于第四层。使用线性控制原理，辅以非线性补偿控制机械手，使学得的机械手动力学知识创建一个在整个工程中高速控制机械手的控制器。模拟结果表明控制器的性能得到了大大提高。
(Abstract: This paper presents a solution to the issue of robot control neural network method. Using a hierarchical neural network (NN) structure to study the characteristics of rigid robot dynamics. For the general category of the mechanical hand, the use of a series of three training feedforward network module, and then these basis functions in real time for the fourth floor. The use of linear control theory, supplemented by non-linear compensation control robot to learn the robot dynamics knowledge to create a project in the entire high-speed control manipulator controller. Simulation results show that the controller performance has been greatly enhanced.)