Simulating lifting motions using artificial neural networks

Abstract

Human motion simulation has never been done accurately. It is a major challenge for engineers and researchers to integrate humans in CAD systems. Human motion simulation has been used in various applications which have been successful. They have increased the effectiveness, safety and efficiency of the application. This dissertation primarily focuses on study of kinematics and posture of an individual while performing a lifting task especially the ones related to manual materials handling. The postures were predicted using Artificial Neural Networks. A set of inputs like target location were fed into the neural networks and the static postures were generated as output. Another network with target locations and time duration of movement where given as inputs to predict joint angles and joint positions. The postures were predicted for two dimensional sagittally symmetric lifts as the number of degrees of freedom is less. The networks were trained using data captured from real human data involving lifting motions. The mean square errors were calculated to validate the accuracy of the prediction.