Towards lifelong visual maps

The typical SLAM mapping system assumes a static environment and constructs a map that is then used without regard for ongoing changes. Most SLAM systems, such as FastSLAM, also require a single connected run to create a map. In this paper we present a system of visual mapping, using only input from a stereo camera, that continually updates an optimized metric map in large indoor spaces with movable objects: people, furniture, partitions, etc. The system can be stopped and restarted at arbitrary disconnected points, is robust to occlusion and localization failures, and efficiently maintains alternative views of a dynamic environment. It operates completely online at a 30 Hz frame rate. A mobile robot existing in a space over a period of time has to deal with a changing environment. Some of these changes are ephemeral and can be filtered, such as moving people. Others are more permanent: objects like furniture are moved, posters change, doors open and close, and more infrequently, walls are torn down or built. The typical SLAM mapping system assumes a static environment and constructs a map, usually in a single continuous run. This map is then enshrined as the ground truth, and used without regard for ongoing changes, with the hope that a robust localization filter will be sufficient for navigation and other tasks.