On SLAM as a Service for Self-Driving Cars

Abstract

Simultaneous Localization And Mapping (SLAM) is one of the common methods used to navigate mobile robots in unknown environments without GNSS. The paper describes the activities from current research projects on how to adapt this method for self-driving cars and how the typical major drawbacks of common SLAM methods can be solved. The ambition are cars sharing their sensor data with a cloud service that aggregates the data into a geo-referenced map, and cars using this map service for localization while driving. The paper presents how the data aggregation works, how to deal with network and timing issues, and how to apply the method for navigation in real urban traffic. The results show a connected car with improved positioning in relation to GNSS and self-driving experiments in public place.