Abstract: This paper deals with the inverse kinematics problem of the PUMA560 robot. Based on six similar BP neural networks, the inverse kinematics solution of the manipulator is obtained. The training parameters are obtained by optimized organization using orthogonal design experimentation (ODE). Simulation results indicate that this method has the advantage of fast learning rate, high precision of identification and better real-time capability.

Key words: inverse kinematics, robot, BP network, orthogonal design experimentation (ODE)