Localization Method for Autonomous Vehicles by using Vertical Elements in the Environment

Abstract

The goal of this work is to provide autonomous cars with an alternative method to estimate their position in the world. It focuses on those situations when the GPS signal gets imprecise, weak or even lost due to environment conditions (long tunnels, dense forests, high buildings around, etc.). By installing two Velodyne VLP-16 LiDAR devices on the top of the car, located on both front corners of the vehicle structure, it is possible to acquire a 3D representation of this environment as a point cloud. This point cloud will be analyzed in order to find thin, vertical elements, such as traffic signals, traffic lights and so on. If these elements have been previously georeferenced, this is, their coordinates have been acquired and saved, the method will try to compare and match the incoming information from the environment to the georeferenced objects, i.e., performing data association. If the detected elements match any of the georeferenced ones, the car will correct an estimate of its position in a global frame.