Learning Robot Dynamic Characteristic with Classification Neural Network

This paper presents a neural network method to solve robot control problem. A classification neural network is used to learn rigid robot dynamic characteristic. To common manipulator training a serial of three-layer forward –feedback module before using, then these radical function is used in four-layer in real time. Based on principle of linear control and compensated with nonlinearity the gained manipulator dynamic knowledge establishes a controller that control manipulator in the whole process. Simulation shows the performance is proved greatly.