Year: 2017
Type: dissertation
Abstract: Localisation of the vehicle is one of the most important part of autonomous driving and can be achieved in many different ways.Most autonomous vehicles today are not able to navigate in complex environments, for instance snow covered roads or heavy rain.This thesis proposes a method for localising a vehicle using LiDAR and Camera fused together with INS and GNSS.The method is intended to work in the complex environments mentioned earlier.The system consists of a vehicle equipped with a 3D scanning LiDAR, a forward looking camera, a combined INS/GNSS unit and a 3D map over the area of interest.The LiDAR and camera are used together to measure distances to points of interest in the 3D map which are then used to triangulate the position of the vehicle.The position from the triangulation is then fused with INS and GNSS in an Ensemble Kalman Filter along with a motion model of the vehicle to estimate the position of the vehicle.It is concluded that the system is promising with a mean translational error of about 40 centimeters which is a good result considering that DGPS has about the same accuracy.
Authors: Thang B. Hoang, Viktor Palmqvist Berntsson
Language: English
Subfield: Aerospace Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: green
Funder: University of Oxford
OpenAlex ID: W2724448693