This paper will describe the current research and technology that is implemented the Keystone rescue team. This year will be the third year that we have competed at the AAAI USAR competition. Our research focus is on developing fully autonomous robots that can maneuver in unstructured environments. All of the work that is being done by our group is computer vision based, utilizing low level image processing (edges, skeletons, regions, etc...) to perform ego-motion estimation, stereo matching and VSLAM (Visual simultaneous localization and mapping). In order to keep our research purely in the computer vision domain, we use cameras as the sole sensor (e.g., no LADAR, shaft encoders, etc...). This allows us to develop robust algorithms using inexpensive (noisy) platforms. Section describes the robots used in the research. Section provides background information on previously approaches to stereo vision, ego-motion and SLAM. Section describes the design and implement of our system. Section explains our approach to ego-motion estimation and section will detail our stereo vision work. A brief discussion of the results as well as directions for future research are in Sec.
展开▼