抄録
This paper describes a method to update a map according to the observation results on unknown obstacles acquired by a mobile robot during its travel. The map is two dimensional on which obstacles are described with polygons or circles. The sensed data obtained by observations with a range-finder consist of sensed points and directions on obstacle surfaces. The map update is subdivided into 3 levels as follows, (1) describe imaginary obstacles onto the map that represent each sensed point, (2) find clusters from sensed points and describe imaginary obstacles that represent their sizes and positions, (3) describe obstacle shapes by estimating them with sensed data. These levels are selectively used according to the amount of data and allowed time to spare. Level 1 and 2 are used in real-time when the robot is moving, and level 3 is conducted when sufficient data and time are available. In order to estimate unknown obstacle shapes with sensed data, an algorism is developed that selects and binds neighbor pairs among sensed points. The algorism can estimate concave or multiple shapes. Experiments using a mobile robot demonstrated utility of each level.
