LiDAR based mesh generation for virtual vehicle testing and environmental simulation

Abstract

Digital twins and light detection and ranging (LiDAR) are significant topics of interest among many industries today, and the automotive industry is one of the major users who greatly benefit by utilizing these technologies. Car manufacturers have used digital twins for the purpose of efficiently testing their products in terms of performance and safety, and LiDAR has been used for example in many of the autonomous driving systems seen today.

The recent advancements on LiDAR technology have made it possible for professional LiDAR devices, such as the GeoSLAM Zeb Horizon used in this study, to produce dense point clouds of a target environment with a range of up to a hundred meters. This along with the increased use of digital twins for testing purposes has created a use case for mapping real world environments for the purpose of them functioning as a digital twin in a simulation environment for virtual testing of vehicles.

This thesis contained an introduction to LiDAR technology and digital twins in the form of a narrative literature review, and furthermore explored the practical possibilities by creating a workflow where a LiDAR scanned point cloud was turned into a 3D mesh model using different tools to see if the reconstructed model of the point cloud is capable of functioning as simulation environment. As a result, a reconstructed 3D mesh model of a point cloud made using Agisoft Metashape was imported into Unreal Engine 5, and it was concluded that with the used LiDAR device it was possible to create a model of a target environment with its physical features being mostly well preserved, especially regarding the ground surface.