Occupancy grid mapping with the use of a forward sonar model by gradient descent

Abstract

Mapping of the environment is one of the key functions of autonomous robots. An efficient mapping algorithm is needed for successful localization and path finding. Maps may be plotted on the basis of the data from various sensors. Cheap and easy-to-install sonars are typically used for this purpose. A sonar-based mapping algorithm based on continuous optimization methods is proposed. This algorithm may be used in real time, and its main advantages over the common methods are more accurate determination of the size of small (smaller than the sonar beam width or comparable to it) obstacles and more reliable detection of narrow passages (e.g., doorways).