Advanced Mapping and Localization for Autonomous Vehicles using OSM

Abstract

Path planning plays an important role in various fields of application and research among robotics and autonomous vehicles. Correct path planning requires accurate vehicle positioning and tracking which are two highly crucial tasks for numerous applications in the field of intelligent transportation systems. Global Positioning System (GPS) is extremely affected by various error sources and accordingly produce inaccurate trajectory data. To make this data useful, it has to be linked to the road network with the aid of map matching. This paper introduces a method for processing Geo-data from OpenStreetMap as an environmental representation and also a method for solving the shortest route problem (route planning) using the bidirectional Dijkstra’s algorithm. Moreover, the point to edge online map matching method is implemented using the data extracted from OpenStreetMap road network and Robot Operating System. These approaches are tested on a "Mitsubishi i-MiEV" fully electric vehicle in terms of performance and quality. The results discuss the quality of the path and local occupancy grid map that is generated. They also show that the matching technique presented is able to keep up with input frequencies up to 100Hz and produce high performing output compared to matching using linear search algorithms.