Manipulator calibration based on PSO-RBF neural network error model

Abstract

In order to improve the positioning accuracy of manipulators of the double-arm rock drill rig, a manipulator calibration method based on radial basis function (RBF) neural network optimized by particle swarm optimization (PSO) is proposed. The MDH method is used to establish the forward kinematics model of the 5-DOF manipulator of the double- arm rock drill rig to solve the problem of parallel joint singularity. Using the output root mean square error of RBF neural network as the fitness function of PSO, the PSO-RBF neural network error model of the manipulator is established. The analysis shows that when the training set is small, the increase of training set can greatly improve the accuracy of model. However, when the training set is large, the increase of training set data is difficult to improve the accuracy of model training due to the complexity of the model. When the training data is enough, the average position error of the manipulator end after calibration using the PSO-RBF neural network calibration method is reduced by 93.66% compared with that before calibration, and the maximum position error of the manipulator end is reduced by 84.6%. It is proved that the PSO- RBF neural network calibration method can effectively improve the positioning accuracy of the manipulator.