인간 추종 로봇의 성능 향상을 위한 인간 위치 평가에 관한 연구

Abstract

A Human tracking robot is a robot that can follow a human by judging the location of the human in a real time. Conventional human tracking robots using a camera or an ultrasonic sensor have a serious problem that the robot can not follow a human because they can not recognize human's position in an inclined or obstacle environment. To solve this problem a human tracking method using GPS was developed. Since the position of the person is recognized by using the GPS only, this method requires the use of expensive and accurate DGPS in order to reduce the position error. Another problem is that it is impossible according to satellite or ionosphere conditions. In terms of practical use of human tracking robot, it is necessary to develop a human tracking method based on an inexpensive and inaccurate GPS. In this case, however, since the position error increases and the position changes severely due to the random nature of the DGPS error, the human tracking performance deteriorates significantly. In this paper, we propose an localization of a human position based on the information of the average human movement and rotation by using inexpensive GPS. This method not only reduces the position error but also smoothes the path of the human. It can also enable the robot to follow the human even when the satellite signal is not received in a short period of time. To do this, we estimate the average amount of traveling distance and rotation of human, and these distance and rotation information are fused with GPS information using extended Kalman filter. This method can improve the human tracking performance of the robot because it can reduce the position error and smoothen the sudden change of the path. In order to evaluate the applicability and possibility of the method in real environment, we present experimental results for various situations using a real robot in a real world.