Fuzzification-back propagation neural network-based model prediction for robotic arm positioning error reduction

Robotic arms are pervasively used in critical manufacturing fields, but its absolute positioning accuracy cannot be controlled because of random errors . Although some researches using back propagation neural network to predict the robotic arm's absolute positioning error have been proved, they suffer from poor convergence and low prediction accuracy in that input parameters contain unavoidable measurement errors. This paper proposed an error prediction model based on fuzzification and back propagation neural network. The rotation angle and direction of robotic joints are employed as training samples for the back propagation neural network, and converted into error contributions using the fuzzification to eliminate the influence of measurement errors in the input parameters. The input parameters are simplified, which enables the training process of the back propagation neural network to be optimized. Experimental results showed that the training time of the model was reduced by two times or more, and the mean square error was decreased by roughly 2.94 %. Meanwhile, the average absolute positioning error of the robotic arm with the prediction model was reduced by 59.22 %. The model can be easily transplanted into embedded systems to provide a methodology for new design of robotic arm error compensators.