Road Network Variation Based on HD Map Analysis for the Simulative Safety Assurance of Automated Vehicles

TL;DR

This paper presents a method to analyze real HD maps and generate synthetic road networks with realistic variations for large-scale automated vehicle simulation testing.

Contribution

It introduces a novel approach to extract parameters from HD maps and generate diverse, realistic road network variations in OpenDRIVE format for automated vehicle testing.

Findings

Parameters extracted from HD maps enable realistic network variations.

Generated road networks improve simulation diversity.

OpenDRIVE-based models facilitate large-scale testing.

Abstract

The validation and verification of automated driving functions (ADFs) is a challenging task on the journey of making those functions available to the public beyond the current research context. Simulation is a valuable building block for scenario-based testing that can help to model traffic situations that are relevant for ADFs. In addition to the surrounding traffic and environment of the ADF under test, the logical description and automated generation of concrete road networks have an important role. We aim to reduce efforts for manual map generation and to improve the automated testing process during development. Hence, this paper proposes a method to analyze real road networks and extract relevant parameters for the variation of synthetic simulation maps that correspond to real-world properties. Consequently, characteristics for inner-city junctions are selected from Here HD map. Then, parameter distributions are determined, analyzed and used to generate variations of road networks in the OpenDRIVE standard. The presented methodology enables efficient road network modeling which can be used for large scale simulations. The developed road network generation tool is publicly available on GitHub.