AttentionMapを用いた注意推定によるアシストの選別

抄録

Today, with the development of robotics and network technologies, robots that can be remotely controlled based on images from cameras have been developed. Since these robots are difficult to operate due to their remote-controlled nature, researches being conducted to reduce the difficulty of operation by implementing operation assist System. However, depending on the timing of the operation assist, it may interfere with the work or induce hazards. Therefore, in this research, based on the assumption that such unwanted assists will occur, aims to improve the reliability of operation assists by eliminating unwanted assists through assist selection. When operate a remote-controlled robot, the operator has to move its line of sight in order to check the camera images. Therefore, we have studied a method to estimate the distribution of attention during operating from eye gaze and to select assists according to the degree of attention. First, we prepared a robot equipped with a control support system in a virtual space, and asked subjects to remotely control the robot. At that time, we selected assistants using the proposed method and evaluated their task completion time and the number of mistakes they made. As a result, it was confirmed that unnecessary assists were eliminated, but there was a tendency that some necessary assists were also eliminated.