Estimating Joint Torques of a Robotic Arm Based on Kinematics Using Neural Networks

This paper presents a methodology to determine robot joint torques based on kinematic quantities by utilizing artificial neural networks (ANN). Due to time series nature of kinematic quantities, a physics-based measure for kinematic quantities has been formulated and implemented to train the neural network. In addition, the effect of the different number of hidden layers in the neural network was compared as well as the effect of different training methods in determining the joint torques. Experimental results confirmed the validity of the proposed method to determine the joint torques with high accuracy.