Development of a semantic map for an unmanned vehicle using a simultaneous localisation and mapping method

Abstract

Introduction: The field of unmanned technologies is rapidly developing and a lot of research is being conducted on the practical application of artificial intelligence algorithms to solve complex problems on the road. The difficulties in the perception of the surrounding world by the machine led to the appearance of special High definition maps. These maps are used to simplify and improve the quality and reliability of other subsystems from the stack of autonomous technologies, such as localization, prediction, navigation and planning modules. In modern literature, there are mainly works on the practical application of such maps, and the process of developing a map remains outside the scope of consideration.The aim of the work is to create a methodology for designing semantic maps for autonomous vehicles with a detailed description of each of the development stages.Materials and methods: The article describes the methodology for creation of HD maps, which includes the stages of data collection using SLAM (Simultaneous localization and mapping) approach, its further processing and the development of the semantics of the road network. The described algorithm is applied in practice to develop the semantic map of Innopolis city area using SLAM approach with LIDAR inertial odometry via smoothing and mapping (LIO-SAM).Results: The main stages of the methodology for creating HD maps for autonomous vehicles have been proposed and investigated. Authors implemented the proposed concept in practice and described in detail the process of creating a semantic map for the Innopolis city area.Conclusions: The proposed methodology can be used for any type of autonomous robots (ground vehicles, unmanned aerial vehicle, water transport) and can be implemented in different road conditions (city, off-road), depending on the information the map should provide for the implementation of the goals and objectives set for the autonomous vehicle.