Case study
Accelerating L4 simulation at scale for a large European automotive OEM
How HERE enables faster, scalable simulation workflows for autonomous driving with AI-driven scene extraction and parametrization.
Industry: Automotive
Region: Global
HERE products used: SceneXtract
Key Results:
68%
less time spent identifying and building scenes
100
distinct real-world driving environments
< 2 days
to sourcing simulation-ready environments
The challenge
As the European OEM advanced its Level 4 (L4) autonomous driving systems, simulation became a critical pillar of its validation strategy. The company needed to generate highly realistic environments and generate a large volumes of scenario variations to test system performance at scale.
However, building these environments presented a significant bottleneck. Traditional approaches to sourcing and preparing HD map data required time-intensive manual effort, limiting the OEM’s ability to quickly iterate across scenarios. Additionally, generating multiple parameter variations on the same base environment further compounded complexity, slowing down simulation throughput and increasing costs.
To meet its goals, the OEM needed a more efficient, scalable way to create simulation-ready environments that could keep pace with the demands of high-volume, high-fidelity testing.
The solution
To achieve the speed and scale required, HERE delivered a tailored solution using SceneXtract, which was built on AWS using Amazon Bedrock and OpenSearch. The OEM identified the characteristics and requirements for ten base scenes, which were used to identify 10 unique variations per base scene, resulting in 100 total simulation-ready environments.
This approach streamlined the simulation workflow and is designed for scale. At the core of the solution was a two-step architecture:
Curated base map library: Using SceneXtract’s AI-powered location search capabilities, HERE created a curated library of high-quality base maps aligned to the simulation needs. These maps were derived from HERE HD Live Map data and preconfigured for immediate use.
On-demand parametrization via API: The OEM could then dynamically modify these base maps through a single API. The parametrization engine enabled semantic changes, such as road attributes, lane types, markings, and traffic elements, without altering underlying geometry. At the same time, a scenario realignment module ensured that OpenSCENARIO files were automatically adapted to the updated maps.
This separation of map sourcing from runtime parametrization was critical. It allowed the OEM to reuse the same base environment while generating hundreds of variations quickly and consistently, eliminating the need to repeatedly rebuild scenes from scratch.
The solution also supported both single and bulk queries, enabling large-scale parameter sweeps in a single request, ideal for high-volume simulation workflows.
The Impact
HERE SceneXtract fundamentally transformed the OEM’s ability to scale simulation for L4 autonomous driving, both in speed and volume.
Scaled from 10 base scenes to 100 simulation-ready assets with ease: With 10 variations for each scene, the OEM significantly increased test coverage without duplicating effort
Cut scene creation time from weeks to hours: Manual map sourcing and preparation was replaced by an API-driven workflow, reducing turnaround time from days or weeks to just hours
Accelerated iteration and validation cycles: Bulk scenario generation enabled teams to test multiple parameter variations in a single run, dramatically speeding up development and feedback loops
Enabled scalable, repeatable simulation workflows: Deterministic, reusable base maps and automated parametrization reduced complexity, lowered costs, and laid the foundation for industrial-scale L4 simulation
By enabling the automaker to scale from a handful of scenes to hundreds of variations in a fraction of the time, HERE SceneXtract delivered a step-change in simulation productivity and accelerated development timelines, while improving the depth and quality of testing.
Why HERE?
Advanced HD mapping and simulation expertise,
enabling precise, real-world representations
Innovative GenAI capabilities
for faster environment creation
Seamless integration with existing simulation tools,
including IPG CarMaker and orchestration platforms