Angular Uncertainty Reduction of Sonar Range Data for a Grid-based Map Building

This study describes an angular uncertainty reduction method of a sonar sensor for building a probabilistic grid map for the autonomous mobile robot. The reliable angle evaluation (RAE) model is based on the information of the hypothesized feature through the association of sonar data that are reflected by the same object. The hypothesized feature is modeled by a set of geometric primitives such as a point, line and arc feature. The RAE model estimates the reliable angle of the detected object on the sonar beam aperture using the geometric relationship between the hypothesized feature and the sonar sensor. Consequently, the RAE model can reduce the angular uncertainty of a sonar sensor. The Bayesian and the orientation model update occupied grids that belong to the region of reliable angle evaluated by the RAE model. The proposed method was tested in a home-like environment using a mobile robot. The resulting grid map shows good quality when it was built in a complex environment with lots of convex and concave corners