Person Following Robot with Vision-based and Sensor Fusion Tracking Algorithm

Abstract

Current demographics show that Japan is experiencing a combination of an aging population and a declining birth rate. Therefore, interest is growing in the potential of human symbiotic robots such as daily life support robots that can care for the aged and young children. Human symbiotic robots require sophisticated capabilities to achieve symbiosis and interaction with humans. It is essential for these robots to understand human intentions, and interact with humans and the environment. We call these which create real value for people and society, technologies, and have developed some home robots and human symbiotic robots (Yoshimi et al., 2004; Matsuhira et al., 2005a; Matsuhira et al., 2005b). The development of these robots is a typical target for human-centric but these technologies are not only applicable for robots but also for all machines that humans use. In the development of human symbiotic robots, we set one of the target criteria as the ability to recognize individuals using vision, and to understand situations in order to provide various real-life services to humans. To realize this target criterion, we think that the principal capabilities required are accurate vision and recognition of specified individuals who are in the vicinity. Moreover, another important capability common to the human symbiotic robot is that of moving safely near humans. In light of the above considerations, we have developed ApriAttendaTM shown in Fig.1, a person following robot that finds a specified person using visual tracking functions and follows him/her while avoiding obstacles (Yoshimi et al., 2006). Person following robots developed until now use various types of cameras for detecting a target person, and some of them use other sensors (Schlegl et al., 1998; Sidenbladh et al., 1999; Kwon et al., 2005; Cielniak et al., 2005; Kobilarov et al., 2006). Our newly developed robot adopts a stereo vision system, and additionally a Laser Range Finder (LRF) is mounted on the robot body to enhance the performance of person following motion. A person following robot has to distinguish the target object from other objects and recognize it by some methods. And the robot has to get the information of the target position, and continue following it quickly so as not to be left behind. At this time, camera information is often used to recognize the target. In addition, in the case of the stereo systems using two or more cameras, the distance information for the person following motion can be obtained. O pe n A cc es s D at ab as e w w w .ite ch on lin e. co m