Modeling and Simulation of Robot Inverse Dynamics Using LSTM-Based Deep Learning Algorithm for Smart Cities and Factories

Abstract

In smart cities and factories, robotic applications require high dexterity and security, which requires precise inverse dynamics model. However, the physical modeling methods cannot model the uncertain factors of the manipulator such as flexibility, joint clearance and friction, etc. As an alternative, artificial intelligence (AI) techniques have become increasingly popular in robotics for smart cities and factories. In this paper, deep learning neural network based on LSTM (Long Short-Term Memory) is adopted to predict the manipulator inverse dynamics. This study aims to summarize the influence of the hyper-parameter settings on model performance and to explore the applicability of the LSTM model to joint torque prediction of multiple degrees of freedom series manipulator. Furthermore, the feasibility of using only joint position as input data for torque prediction is verified. Simulation result has shown that, for the proposed deep learning architecture, the effects of the number of maximum epochs on model performance should be prioritized. The effects of the number of hidden nodes on model performance are limited, while prediction accuracy will deteriorate as the number of hidden layers increases. It is proved that it is feasible to predict inverse dynamics when input data is joint position only. The experimental results show that the training time increases with the increase of hidden layers, neurons and epochs.