A vision based robotics system capable of following a person executing a mission has many applications such as search, rescue or combat in previously unvisited locations. Its main advantage over other modes of person following is its passivity in the sense that it does not transmit or receive detectable signals. However, vision based robotic person following in outdoor environments is a very challenging problem due to extreme variations in the lighting conditions and its effect of the color of person’s outfit, as well as the effect of the reflected colored lights from buildings, trees, etc.
As the first phase of a project sponsored by CCAT and SSC, we have developed a system that to a large extend addresses the difficulties mentioned above by controlling various camera and robot parameters. The research and development effort has resulted in a robotic system that is capable of following a person wearing a single color shirt in several environments both indoors and outdoors. The robot has been equipped with an onboard computer, vision system and various other components. Novel algorithms and software have been developed as part of this project for various image processing and control functions. However, due to the short duration of the first phase, the existing system is not robust and further improvements are planned to deal with the robustness issue, multi-colored outfits, etc.
