An Approach to Systematic Data Acquisition and Data-Driven Simulation for the Safety Testing of Automated Driving Functions

TL;DR

The paper addresses the challenge of validating automated driving functions in complex open environments where real-world data are scarce. It proposes a data-driven pipeline that collects heterogeneous traffic data (aerial, infrastructure, and vehicle-based) and converts it into map-referenced trajectories to parameterize the driving-behavior model framework, implemented in a $\text{Wiedemann99}$-based simulation within PTV Vissim and a Carla-based SUT digital twin. A probabilistic, likelihood-based optimization fits observed mean velocities of cars and trucks, enabling stochastic validation and generation of diverse traffic dynamics for virtual testing, including near-miss scenarios defined for stochastic risk assessment. The work highlights the need for standardized international data formats (OpenDRIVE/OpenSCENARIO/OpenLABEL) and long-term data collection to realize scalable, quantitative virtual validation for SAE Level 4/5 automated driving systems.

Abstract

With growing complexity and criticality of automated driving functions in road traffic and their operational design domains (ODD), there is increasing demand for covering significant proportions of development, validation, and verification in virtual environments and through simulation models. If, however, simulations are meant not only to augment real-world experiments, but to replace them, quantitative approaches are required that measure to what degree and under which preconditions simulation models adequately represent reality, and thus, using their results accordingly. Especially in R&D areas related to the safety impact of the "open world", there is a significant shortage of real-world data to parameterize and/or validate simulations - especially with respect to the behavior of human traffic participants, whom automated driving functions will meet in mixed traffic. We present an approach to systematically acquire data in public traffic by heterogeneous means, transform it into a unified representation, and use it to automatically parameterize traffic behavior models for use in data-driven virtual validation of automated driving functions.

Figures (5)

• Figure 1: Samples of the aerial image traffic data acquisition method, based on images recorded by a Flight Design CTSW ultralight aircraft between the cities of Karlsruhe and Rastatt.
• Figure 2: A point cloud from the infrastructure recording after applying the processing pipeline. Background points are colored in blue and points belonging to an object have a different color based on the tracked object.
• Figure 3: Comparison between the histogram of the extracted mean velocities from our recorded data (trucks: green, cars: blue) and the respective density estimations determined from the simulated mean velocities (trucks: orange, cars: black).
• Figure 4: Comparison of the respective 100 resulting $\mu$ and $\sigma$, of each individual estimation run. The parameters for the truck type are depicted in green and car type in blue. Further, initial guesses for each estimation run are shown in gray.
• Figure 5: Variation of the minimal observed mean (top) and minimal observed minimum (bottom) time to collision (TTC) values when varying Wiedemann99 parameters for a subset of car agents (different colors encode different parameter variations). Only the parameters cc1 (orange) and cc3 (red) show significant influences.