Application of ANN in Identification of Inertial Parameters of End-Effector of Robot

Abstract

Identification of inertial parameters of manipulator of a robot is an important basic problem in robot dynamical research. Algorithm and mathematical model of inertial parameters identification of robot are analyzed in this paper. General rules and advantages of application of artificial neural network in identification of parameters of system are investigated also. A novel method different from the traditional ANN problems is developed. The structure and the weights of the network constructed have special physical meanings in this method. Difficulty in obtaining the training samples, which exists in the traditional ANN for inertial parameters estimation, is resolved. This method is simple, direct and less in operand, which can be applied on the end-effector of a robot on line to estimate its inertial parameters based on the signal of sensors. Feasibility and effectiveness of this method are proved by the results of experiment carried out in a PUMA 562 robot.