Inertia parameter identification of robot arm based on BP neural network

Abstract

The modeling and controlling of robot dynamics are two important fields in the robotics. Modeling is the precondition of controlling. Accurate model parameters obtained can improve the control precision. In the paper, the dynamic model of a robot arm is built with the Newton-Euler method and transformed into linear equations about inerta parameters for identification By operating the robot arm, the system input and output data can be abstracted and a BP neural network is to create. The 10 inertia parameters of every connecting rod are regarded as the weights of the neural network. The errors of output torques between the original system and the neural network are used to adjust the weights. Finally, the results of inertia parameters identification are obtained. Then take a two degree-of-freedom robot arm as an example. The simulation result verifies the validity of inertia parameter identification based on neural network.