Localization and On-Line Map Building for an Autonomous Mobile Robot

Two basic tasks for an autonomous mobile robot are tackled in this paper: localization and on-line map building. For localization the robot uses a laser radar with 720 data points on a full circle with a resolution of a few mm. Localization is done by correlating radar pictures from a point P0 to a point P1 nearby. The radar pictures are transformed into angle histograms and correlated to give the relative rotation Δφ of both. The maximum angle γ in the first histogram shows the main direction in the environment The histograms are turned into this direction. Point histograms correlated give translations Δx and Δy between the poses P0 and P1. Rotating backwards by γ the pose P1 is given in the coordinate system at P0. They are reference points. Map building uses an overlay of radar pictures from an obstacle avoiding radar to find points of interest (POI) in the near vicinity of the momentary position. They will form reference points too. A backtracking algorithm forms a topological graph of reference points exhausting the environment. The robot starts into an unknown environment and knows when it has explored it. The topological graph and the radar pictures at the nodes represent the environment.