Feature based map building using sparse sonar data

A new feature based map building model that uses only the footprints of sparse sonar data has been developed and implemented. An arc feature association model was developed, which associates two sonar footprints into an arc feature. The association model provides information on the regions of center points of the arc feature. The information makes it possible to decide weather the two sonar footprints are associated with a line, point or arc. Lines, points, or arc features are extracted from more than two independent sonar footprints using the arc feature association model. We also propose a method that estimates the position uncertainties of the extracted features by considering both the pose uncertainty of the robot and the measurement uncertainty of the sonar sensor. This pose uncertainty is also used in the arc feature association process of the sonar footprints. The proposed method has been tested in a real home environment with a mobile robot.