Geo-referencing Autonomous Vehicles Using LoD2 and HD Maps: Performance Assessment in Simulated Urban Environments

Publikation: KonferenzbeitragPaperForschungPeer-Review

Forschungs-netzwerk anzeigen

Details

OriginalspracheEnglisch
Seiten915-922
Seitenumfang8
PublikationsstatusVeröffentlicht - 14 Juli 2025

Abstract

Autonomous vehicles (AVs) require accurate global pose estimation to operate effectively. A common approach involves utilizing perception sensors to extract environmental features which are used to geo-reference the vehicle with pre-defined maps. High Definition (HD) maps are frequently used for this purpose due to their detailed feature sets. However, the use of HD maps presents challenges as they are not frequently unavailable and their custom generation involves considerable complexity and cost. Conversely, Level of Detail 2 (LoD2) maps are freely available for numerous cities and are regularly updated, hence they can offer a potential solution. However, due to their geometric simplifications, the applicability of LoD2 maps for AV pose estimation remains uncertain. In this study, we investigate the impact of these simplifications and assess the suitability of LoD2 maps for AV pose estimation. We perform a comparative analysis between HD and LoD2 maps in a simulated CARLA environment, employing an Error State Kalman Filter (ESKF) to estimate the position, velocity, and orientation of an AV. We showcase our results using ideal sensors to isolate the effects of LoD2 maps, as well as realistic sensors to evaluate their performance in real-world scenarios.

ASJC Scopus Sachgebiete

Zitieren

Geo-referencing Autonomous Vehicles Using LoD2 and HD Maps: Performance Assessment in Simulated Urban Environments. / Wahbah, Mohamad; Ramme, Lukas; Vogel, Sören et al.
2025. 915-922.

Publikation: KonferenzbeitragPaperForschungPeer-Review

Download
@conference{421fc1f3e8ff4dde9a6dcd7332b7c001,
title = "Geo-referencing Autonomous Vehicles Using LoD2 and HD Maps: Performance Assessment in Simulated Urban Environments",
abstract = "Autonomous vehicles (AVs) require accurate global pose estimation to operate effectively. A common approach involves utilizing perception sensors to extract environmental features which are used to geo-reference the vehicle with pre-defined maps. High Definition (HD) maps are frequently used for this purpose due to their detailed feature sets. However, the use of HD maps presents challenges as they are not frequently unavailable and their custom generation involves considerable complexity and cost. Conversely, Level of Detail 2 (LoD2) maps are freely available for numerous cities and are regularly updated, hence they can offer a potential solution. However, due to their geometric simplifications, the applicability of LoD2 maps for AV pose estimation remains uncertain. In this study, we investigate the impact of these simplifications and assess the suitability of LoD2 maps for AV pose estimation. We perform a comparative analysis between HD and LoD2 maps in a simulated CARLA environment, employing an Error State Kalman Filter (ESKF) to estimate the position, velocity, and orientation of an AV. We showcase our results using ideal sensors to isolate the effects of LoD2 maps, as well as realistic sensors to evaluate their performance in real-world scenarios.",
keywords = "Digital Maps, Geo-referencing, LIDAR, Simulation",
author = "Mohamad Wahbah and Lukas Ramme and S{\"o}ren Vogel and Ingo Neumann and Hamza Alkhatib",
note = "Publisher Copyright: Copyright {\textcopyright} 2025 Mohamad Wahbah et al.",
year = "2025",
month = jul,
day = "14",
doi = "10.5194/isprs-annals-X-G-2025-915-2025",
language = "English",
pages = "915--922",

}

Download

TY - CONF

T1 - Geo-referencing Autonomous Vehicles Using LoD2 and HD Maps: Performance Assessment in Simulated Urban Environments

AU - Wahbah, Mohamad

AU - Ramme, Lukas

AU - Vogel, Sören

AU - Neumann, Ingo

AU - Alkhatib, Hamza

N1 - Publisher Copyright: Copyright © 2025 Mohamad Wahbah et al.

PY - 2025/7/14

Y1 - 2025/7/14

N2 - Autonomous vehicles (AVs) require accurate global pose estimation to operate effectively. A common approach involves utilizing perception sensors to extract environmental features which are used to geo-reference the vehicle with pre-defined maps. High Definition (HD) maps are frequently used for this purpose due to their detailed feature sets. However, the use of HD maps presents challenges as they are not frequently unavailable and their custom generation involves considerable complexity and cost. Conversely, Level of Detail 2 (LoD2) maps are freely available for numerous cities and are regularly updated, hence they can offer a potential solution. However, due to their geometric simplifications, the applicability of LoD2 maps for AV pose estimation remains uncertain. In this study, we investigate the impact of these simplifications and assess the suitability of LoD2 maps for AV pose estimation. We perform a comparative analysis between HD and LoD2 maps in a simulated CARLA environment, employing an Error State Kalman Filter (ESKF) to estimate the position, velocity, and orientation of an AV. We showcase our results using ideal sensors to isolate the effects of LoD2 maps, as well as realistic sensors to evaluate their performance in real-world scenarios.

AB - Autonomous vehicles (AVs) require accurate global pose estimation to operate effectively. A common approach involves utilizing perception sensors to extract environmental features which are used to geo-reference the vehicle with pre-defined maps. High Definition (HD) maps are frequently used for this purpose due to their detailed feature sets. However, the use of HD maps presents challenges as they are not frequently unavailable and their custom generation involves considerable complexity and cost. Conversely, Level of Detail 2 (LoD2) maps are freely available for numerous cities and are regularly updated, hence they can offer a potential solution. However, due to their geometric simplifications, the applicability of LoD2 maps for AV pose estimation remains uncertain. In this study, we investigate the impact of these simplifications and assess the suitability of LoD2 maps for AV pose estimation. We perform a comparative analysis between HD and LoD2 maps in a simulated CARLA environment, employing an Error State Kalman Filter (ESKF) to estimate the position, velocity, and orientation of an AV. We showcase our results using ideal sensors to isolate the effects of LoD2 maps, as well as realistic sensors to evaluate their performance in real-world scenarios.

KW - Digital Maps

KW - Geo-referencing

KW - LIDAR

KW - Simulation

UR - http://www.scopus.com/inward/record.url?scp=105013239857&partnerID=8YFLogxK

U2 - 10.5194/isprs-annals-X-G-2025-915-2025

DO - 10.5194/isprs-annals-X-G-2025-915-2025

M3 - Paper

SP - 915

EP - 922

ER -

Von denselben Autoren