Improved occupancy grids for map

Occupancy grids are a probabilistic method for fusing multiple sensor readings into surface maps of the environment. Although the underlying theory has been understood for many years, the intricacies of applying it to realtime sensor interpretation have been neglected. This paper analyzes how refined sensor models (including specularity models) and assumptions about independence are crucial issues for occupancy grid interpretation. Using this analysis, the MURIEL method for occupancy grid update is developed. Experiments show how it can dramatically improve the fidelity of occupancy grid map-making in specular and realtime environments.