Robot Evidence Grids

The evidence grid representation was formulated at the CMU Mobile Robot Laboratory in 1983 to turn wide angle range measurements from cheap mobile robot-mounted sonar sensors into detailed spatial maps. It accumulates diffuse evidence about the occupancy of a grid of small volumes of nearby space from individual sensor readings into increasingly confident and detailed maps of a robot’s surroundings. It worked surprisingly well in first implementation for sonar navigation in cluttered rooms. In the past decade its use has been extended to range measurements from stereoscopic vision and other sensors, sonar in very difficult specular environments, and other contexts. The most dramatic extension yet, from 2D grid maps with thousands of cells to 3D grids with millions, is underway. This paper presents the mathematical and probabilistic framework we now use for evidence grids. It gives the history of the grid representation, and its relation to other spatial modeling approaches. It discusses earlier formulations and their limitations, and documents several extensions. A list of open issues and research topics is then presented, followed by a literature survey.