STRADVISION and aiMotive Advance ADAS Development with Integrated Perception and Neural Simulation Workflow

STRADVISION and aiMotive announced the successful completion of a joint proof-of-concept on May 28, showcasing the integration of camera perception technology with ISO 26262-certified neural simulation within a unified ADAS development workflow. The collaboration connects perception and simulation processes to improve scenario coverage while supporting a more efficient path toward advanced driver assistance systems and autonomous driving deployment.

Integrated Workflow for Enhanced Scenario Coverage

The proof-of-concept combines STRADVISION’s SVNet perception platform with aiMotive’s simulation technologies to create a streamlined development pipeline. SVNet provides operational design domain (ODD)-aware interpretation of road environments and extracts relevant driving scenarios from recordings collected in South Korea. These extracted scenarios form the foundation for generating simulation-ready environments that can be used throughout the ADAS validation process.

Neural Reconstruction Enables High-Fidelity Virtual Environments

After scenario extraction, aiMotive’s World Extractor applies neural reconstruction techniques to convert perception-derived scenarios and raw sensor data into highly detailed three-dimensional environments. Using Gaussian Splatting technology, the system creates realistic virtual scenes and produces synthetic sensor data that remains visually and functionally indistinguishable from the original recordings. This approach helps bridge the gap between real-world testing and virtual validation.

Scalable Synthetic Data Generation and Edge-Case Expansion

The synthetic datasets generated through the workflow are produced using aiSim, enabling extensive testing opportunities across a broad range of driving situations. aiFab further enhances coverage by creating scalable scenario variations that target rare and complex edge cases often difficult to capture through conventional road testing. In addition, developers can introduce a diverse range of 3D assets into the environment, allowing the creation of virtually unlimited scene variations for validation and development activities.

Key Components of the Integrated Development Pipeline

• SVNet performs ODD-aware perception and scenario extraction from road recordings.
• World Extractor reconstructs detailed 3D environments using neural reconstruction techniques.
• Gaussian Splatting enables realistic synthetic sensor data generation.
• aiSim produces scalable synthetic datasets for validation activities.
• aiFab generates edge-case variations and additional scene diversity.
• Cloud infrastructure supports large-scale deployment and validation.

Cloud-Scale Validation Supports ADAS Deployment

The complete workflow was validated at scale using cloud infrastructure, demonstrating its ability to handle large volumes of data and simulation tasks efficiently. By linking perception outputs directly with neural simulation and synthetic data generation, the solution offers a comprehensive framework for accelerating ADAS and autonomous driving development while improving testing coverage across increasingly complex driving environments.

Frequently Asked Questions

What is the main objective of the STRADVISION and aiMotive collaboration?
The primary objective is to integrate camera perception and ISO 26262-certified neural simulation into a single ADAS development workflow. This approach improves scenario coverage, enables efficient synthetic data generation, and supports the validation of advanced driving functions. By combining perception-derived scenario extraction with scalable simulation technologies, developers can test a broader range of driving conditions, including rare edge cases, while reducing dependence on extensive real-world data collection.

How does the integrated workflow improve ADAS testing and validation?
The workflow enhances testing by transforming real-world driving recordings into detailed virtual environments and generating synthetic sensor data for large-scale validation. Through aiSim and aiFab, developers can create numerous scenario variations and edge cases that may be difficult to capture on public roads. This expanded coverage allows more comprehensive evaluation of ADAS and autonomous driving systems while supporting scalable cloud-based development and verification processes.