Removing undesired effects of mass/inertia on transparency using Artificial Neural Networks in a haptic mechanism

Abstract

In this paper, Artificial Neural Networks (ANN) has been used to identify the dynamics of robots used in haptic and master slave devices in order to improve transparency. In haptic and master slave devices, transparency depends on some factors such as robot's mass and inertia, gravitational forces and friction [1]. In such systems, mass and inertia of the robot has an undesirable effect on the system outputs, which should be neutralized for improved transparency. The main purpose of this paper introducting a method to neutralize the undesirable effects of mass and inertia of the robot. A recurrent multilayer perceptron (RMLP) is used in a way that the inputs and outputs of the neural network are, respectively, the outputs and inputs of the robot mechanism. Hence, the desired outputs of the mechanism can be given to the neural network as inputs and corresponding required inputs of the robot mechanism can be obtained from the network's output. With this method it is possible to eliminate the undesired influence of mass and inertia on the robot dynamics. The results are compared with the simulations. This comparison shows the effectiveness of using recurrent neural network to achieve this goal.