Learning occupancy grids with forward models

S. Thrun
Proceedings 2001 IEEE/RSJ International Conference on Intelligent Robots and Systems. Expanding the Societal Role of Robotics in the the Next Millennium (Cat. No.01CH37180)  
This paper presents a new way to acquire occupancy grid maps with mobile robots. Virtually all existing occupancy grid mapping algorithms decompose the highdimensional mapping problem into a collection of onedimensional problems, where the occupancy of each grid cell is estimated independently of others. This induces conflicts that can lead to inconsistent maps. This paper shows how to solve the mapping problem in the original, highdimensional space, thereby maintaining all dependencies between
more » ... neighboring cells. As a result, maps generated by our approach are often more accurate than those generated using traditional techniques. Our approach relies on a rigorous statistical formulation of the mapping problem using forward models. It employs the expectation maximization algorithm for estimating maps, and a Laplacian approximation to determine uncertainty.
doi:10.1109/iros.2001.977219 dblp:conf/iros/Thrun01 fatcat:vvn6xb3yaref3fq6yntxo53ycq