Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automatic Odometry-Less OpenDRIVE Generation From Sparse Point Clouds

Leon Eisemann, Johannes Maucher

TL;DR

This work tackles the challenge of generating high-fidelity road representations for automated driving simulation without relying on odometry or extensive sensor setups. It introduces a complete pipeline that derives OpenDRIVE road geometry from sparse LiDAR point clouds by extracting lane markings, building 3D lane representations, and exporting a consistent map for simulation. Quantitative results show centimeter-level lateral accuracy and reproducibility across drives, with highway-scale reconstructions achieving close alignment to HD maps, while qualitative simulations demonstrate realistic traffic scenarios. The approach offers a versatile, sensor-agnostic pathway to HD-map-like road representations suitable for rigorous evaluation of driving functions in simulation, with future work on handling complex lane topologies and broader sensor integration.

Abstract

High-resolution road representations are a key factor for the success of (highly) automated driving functions. These representations, for example, high-definition (HD) maps, contain accurate information on a multitude of factors, among others: road geometry, lane information, and traffic signs. Through the growing complexity and functionality of automated driving functions, also the requirements on testing and evaluation grow continuously. This leads to an increasing interest in virtual test drives for evaluation purposes. As roads play a crucial role in traffic flow, accurate real-world representations are needed, especially when deriving realistic driving behavior data. This paper proposes a novel approach to generate realistic road representations based solely on point cloud information, independent of the LiDAR sensor, mounting position, and without the need for odometry data, multi-sensor fusion, machine learning, or highly-accurate calibration. As the primary use case is simulation, we use the OpenDRIVE format for evaluation.

Automatic Odometry-Less OpenDRIVE Generation From Sparse Point Clouds

TL;DR

This work tackles the challenge of generating high-fidelity road representations for automated driving simulation without relying on odometry or extensive sensor setups. It introduces a complete pipeline that derives OpenDRIVE road geometry from sparse LiDAR point clouds by extracting lane markings, building 3D lane representations, and exporting a consistent map for simulation. Quantitative results show centimeter-level lateral accuracy and reproducibility across drives, with highway-scale reconstructions achieving close alignment to HD maps, while qualitative simulations demonstrate realistic traffic scenarios. The approach offers a versatile, sensor-agnostic pathway to HD-map-like road representations suitable for rigorous evaluation of driving functions in simulation, with future work on handling complex lane topologies and broader sensor integration.

Abstract

High-resolution road representations are a key factor for the success of (highly) automated driving functions. These representations, for example, high-definition (HD) maps, contain accurate information on a multitude of factors, among others: road geometry, lane information, and traffic signs. Through the growing complexity and functionality of automated driving functions, also the requirements on testing and evaluation grow continuously. This leads to an increasing interest in virtual test drives for evaluation purposes. As roads play a crucial role in traffic flow, accurate real-world representations are needed, especially when deriving realistic driving behavior data. This paper proposes a novel approach to generate realistic road representations based solely on point cloud information, independent of the LiDAR sensor, mounting position, and without the need for odometry data, multi-sensor fusion, machine learning, or highly-accurate calibration. As the primary use case is simulation, we use the OpenDRIVE format for evaluation.
Paper Structure (15 sections, 1 equation, 4 figures, 2 tables)

This paper contains 15 sections, 1 equation, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Background
  3. Related Work
  4. Map Format
  5. Method
  6. Data & Preprocessing
  7. Line Marking Extraction
  8. Generation of Lane Marking
  9. Combining Candidate lines
  10. Calculation of Reference Line
  11. Export of OpenDRIVE
  12. Result and Discussion
  13. Quantitative Analysis
  14. Qualitative Analysis
  15. Conclusion and Future Work

Figures (4)

  • Figure 1: General overview of our proposed method. We first build an overall point cloud of the road, filter the respective line markings, generate the respective lanes based on the markings, and in the last step, export into OpenDRIVE. Hereby, the proposed method solely uses point cloud information and exploits geometric properties of lane markings.
  • Figure 2: Overview over lane generation and positional calculation. Lane marking point clouds are displayed in blue, and respective estimated centers in orange. The directional vectors calculated by Eq. \ref{['eq:dir-vec-stabilization']} are shown in shades of green and derived normal-vectors in red. The resulting reference line is displayed in purple.
  • Figure 3: Calculation of the GlobalLookup, as described in Sec. \ref{['subsec:calc-ref-line']}. The relative position is sorted through line ids in the relative positional lookups.
  • Figure 4: Example excerpts of one full drive for qualitative comparison. The respective camera frame and LiDAR scan from the original recording are referenced in each image. The resulting OpenDRIVE is displayed as esmini esmini-environment-2023 rendering, including traffic participants from generated OpenSCENARIOasam-openscenario.