This paper describes a method of making an occupancy grid map through the combined use of DGPS and scan matching. In outdoor environments such as city areas, high-accuracy localization is required for autonomous navigation. Scan matching with a laser scanner and an occupancy grid map consisting of precise structure information on the environment is one of the most accurate localization methods. However, mismatching on the map sometimes occurs, resulting in the robot losing its own position. Although a GPS device, an absolute positioning device, is valid for estimating position and attitude to a certain degree of accuracy, GPS often obtains erroneous positions for multipath problems which occur around tall buildings. In order to estimate the position and attitude of robots more stably, the authors have developed a method of making an occupancy grid map, which corresponds to DGPS directions and has an accurate shape, by using of some accurate DGPS measurement points and the SLAM method. In autonomous navigation, the robot trajectory is estimated using the particle filter method, evaluation and resampling are done using the two ways mentioned above, and attitude is calculated using DGPS measurement points and the result of scan matching. In this paper, the performance of the map-making method and localization method for autonomous navigation is shown through experiments which are evaluated as to the accuracy of the map in an actual environment.
Multivehicle Cooperative Local Mapping: A Methodology Based on Occupancy Grid Map Merging