Generating Human-Like Motion for Arm Robots Using Element Description Method

Abstract

In recent years, there has been significant progress in robotics development, including the creation of humanoid robots. However, when it comes to the motion of the robot's working arm, it is crucial to achieve what is commonly referred to as human-like movement. Failure to achieve this can lead to feelings of fear or discomfort when observing the robot. In order to address this issue, research has been conducted on generating human-like movement by applying machine learning based on human motion. Furthermore, a multi-class classification model is identified using a system identification method called Element Description Method (EDM) to select the most human-like motion from multiple motion plans obtained through inverse kinematics. To improve the accuracy of EDM, actual human joint angles are used as training data, along with questionnaire results on movements considered to be human-like when performed in positions beyond the reach of humans. The generated learning model is then validated for its accuracy, and finally, a comparison is made to evaluate the differences with conventional machine learning methods.