How Korea Testing Laboratory masters scenario generation in
Korea with AVL Dynamic Ground Truth™ System and ADAS/AD
Big Data and Analytics Platform
“To enhance autonomous driving
tests, we used AVL’s solutions to
automate scenario generation and
post-collection processes, which
reduces costs and improves safety
assessment through real data-
derived scenario tagging.”
Generating Logical Scenarios
with Parameters Captured Through
Real-World Data Collection
he Korea Testing Laboratory (KTL) is Korea’s only public
organi^ation in tLe testing certification sector, wLicL
was estaFlisLed to eƾciently sYTTort tLe testing and
evaluation of innovative technology outcomes.
This also applies to the automated driving functions (ADAS
and AD) sector. To ensure public safety, KTL’s goal is to create
a scenario database that can be applied in various forms
of safety validation such as model/software-in-the-loop or
vehicle-in-the-loop.
Deploying AVL supplied technology enables road recordings,
which can be converted to the desired scenario descriptions
semi-automatically. Using standardized formats (ASAM
OpenDRIVE, OpenSCENARIO, and OSI) KTL is able to use these
scenarios in their data pipeline. The tools provided by AVL
comprise a dynamic ground truth measurement system that
is easily mounted and calibrated to the majority of vehicles, a
data analytics platform, and a powerful perception software.
The auto-tagger function within this toolchain allows KTL
engineers to swiJtly refine and discover tLe reUYired scenarios
for their future testing.
The collected data is used for several purposes: KTL analyzes
scenarios within recorded data and evaluates different
KPIs that occur within this scenario. KTL also uses the
parameters of scenarios detected by the auto-tagger to build
a database within AVL SCENIUS™. This database is used to
generate knowledge about the operational design domain.
In addition, test plans can be generated, which are executed
in Model.CONNECT™ or any ASAM compliant tool that is
commercially available or open source.
-n TreTaration Jor tLe 92 )') certification oJ 7%) 0evel
veLicles, /80 is worOing closely witL %:0 to define TrocedYres
to meet all regulatory requirements and use the AVL Analytics
Engine to generate the required reports as automatically as
possible.
HONGSEOK LEE
Senior Researcher ADAS
Korean Testing Laboratory
2024