STRADVISION and aiMotive Combine Perception-Driven Scenario Understanding and Scalable Simulation for ADAS Validation
Joint proof-of-concept demonstrates a real-world-to-simulation pipeline that transforms vehicle fleet data into high-fidelity synthetic environments – indistinguishable from real-world sensor data – validating SVNet perception outputs in real time on cloud infrastructure.
SEOUL, South Korea and BUDAPEST, Hungary, May 28, 2026 /PRNewswire/ -- STRADVISION and aiMotive today announced the results of a joint proof-of-concept demonstrating how production-proven camera perception and ISO 26262-certified neural simulation can work together in an integrated ADAS development workflow.
The collaboration addresses a core challenge in scalable ADAS development: how to convert real-world fleet recordings into hyper-realistic, simulation-ready synthetic environments at scale. STRADVISION contributed its SVNet perception platform – with more than five million cumulative production units deployed globally – to provide ODD-aware interpretation and scenario extraction from Korean road recordings, enabling the identification and structuring of perception-critical driving scenarios for scalable validation workflows. aiMotive's World Extractor then applies neural reconstruction to transform these perception-derived scenarios and raw data into detailed 3D environments using Gaussian Splatting, generating synthetic sensor data that is indistinguishable from the original footage.
The resulting synthetic datasets were produced using aiSim, the world's first ISO 26262 ASIL-D-certified automotive simulator. aiFab can generate diverse scenario variations at scale, covering complex and hard-to-capture edge cases that real-world data collection cannot capture. Moreover, a wide variety of 3D assets – be it dynamic actors, such as vehicles or pedestrians not present in the original recording; or static assets, such as road furniture and traffic signs – can be added to the scenario, to create an infinite number of different scenes. The full pipeline, from raw sensor data ingestion through neural reconstruction, scenario generation, and synthetic data export, was validated running at scale on cloud infrastructure.
The integration establishes a feedback loop between real-world perception and simulation, improving scenario coverage and contributing to more efficient and reliable deployment of ADAS and autonomous driving systems. This feedback loop minimizes the gap between field testing and simulation-based validation – without requiring manual 3D environment creation.
For STRADVISION, the integration unlocks a scalable workflow for transforming proprietary fleet recordings into simulation-ready assets generated within an ASIL-D-certified simulation environment..
"Real-world driving data alone is no longer sufficient to scale validation for next-generation ADAS systems," said Insu Kim, Head of STRADVISION's Data Innovation Center. "Through this collaboration, we demonstrated how perception-driven understanding of complex road scenarios can be transformed into scalable simulation workflows, helping close the gap between field operation and virtual validation."
"We, at aiMotive, strongly believe that safe automated driving requires extensive virtual validation. This project provides proof of how two like-minded and agile companies can build and deploy an efficient, high-quality neural simulation pipeline for an end-to-end automated driving software," said Szabolcs Jánky, SVP of Product Strategy, aiMotive.
The collaboration lays the groundwork for broader integration between perception-driven scenario understanding and scalable simulation-based validation workflows.
For more information about STRADVISION visit: https://stradvision.com
For more information about aiMotive, visit aimotive.com
Contact:
Bence Boda
Marketing Director, aiMotive
[email protected]
SOURCE aiMotive
Share this article