AI-Powered Test Scenario Generation, Validation & Homologation for Safer Autonomous Vehicles
Validate and certify — safer, smarter, and faster with AI. Search for complex test scenarios using natural language queries. Let AI generate detailed scenario descriptions, identify missing edge cases, and simulate rare driving events automatically.
Kanagawa Institute of Technology
Located in Japan, the Kanagawa Institute of Technology has a rich history of over two decades in advancing vehicle systems and autonomous driving technologies. We deeply value our partnership with their forward-thinking department and their significant contributions to the automotive industry.
Benefits of AI, LLM & RAG in Autonomous Vehicle Development & Testing
Natural Language Search for Driving Scenarios: Ask questions like “Which scenarios are existing for a pedestrian crossing in bad weather?” and get AI-ranked results, complete with metadata, references, and links to source and videos.
Precision-Driven Scenario Discovery: AI identifies critical gaps in autonomous driving data, ensuring accurate and reliable test scenarios for safer self-driving technology.
Generate Edge Cases Automatically: AI and predictive models suggest unseen edge cases based on your data gaps—so you can simulate rare and safety-critical events proactively.
Simulation Tool and Test Compatibility: AI supports OpenSCENARIO and connects with external simulation models such as IPG CarMaker or PTV Vissim, making your scenarios ready for simulation.
See AI-Powered Semantic Incident Discovery in Action
Your Questions Answered
It saves time, reduces manual effort, and improves safety. The system automatically finds missing scenarios, completes data gaps, and guides you through testing more efficiently.
How does this reduce risk in autonomous driving?
By spotting what’s missing before it becomes a problem. The AI identifies edge cases, near misses, and safety-relevant incidents through intelligent data retrieval and contextual data analysis—then turns them into test scenarios you can simulate and solve.
Will we get help with setup and training?
Yes. We guide your team through setup, integration, and use cases — so you get value from day one. Hands-on training is included.
Can I save real-world driving data as simulation files and quickly find them again?
Yes, AI with Vision-Language Models (VLMs) automatically converts your recorded drives into simulation files, embeds them into the knowledge base, and makes them discoverable in natural language through our built-in RAG system.
Can the system grow with our data and testing needs?
Definitely. The platform is designed to handle increasing complexity, new data formats, and more scenarios as your testing scales.
How does the system ensure no risks are missed?
The AI constantly checks for gaps in your scenario database and flags areas that need new simulation, lab testing, or real-world driving data. It supports thorough, traceable validation.
Is support for the homologation of L3 and L4 provided?
Yes, through targeted test scenarios and objective evaluations, the homologation of L3 and L4 can be specifically supported — also in collaboration with additional partners like DEKRA and ASVIN.
Can I automatically classify standard maneuvers from recorded drives?
Yes, using automatic trajectory generation and AI methods with Vision-Language Models (VLMs), the system can classify maneuvers like lane changes and left or right turns.
Want to know more?
We're here to help! If you need more information or have any questions, feel free to reach out.
AI Solutions for Our Customers
Explore how our AI solutions have transformed industrial and business processes by enhancing efficiency, reducing manual processes, and driving innovation for our customers.
Start Using Your Knowledge as Competitive Advantage for Your Business
From use case analysis and technology studies to full implementation and integration into your existing IT landscape, EDI offers comprehensive services, including maintenance and hosting, to unlock the full potential of AI for your business.