IAV Mela
To validate the driving functions of all levels of autonomy, you must consider every possible scenario in the specific ODD, which is a challenge.
Validation of autonomous vehicles: challenges and strategies
In the field of autonomous vehicles and advanced driver assistance systems (ADAS), the real challenge is to prepare and validate every possible scenario. To ensure safety, manufacturers apply comprehensive risk management strategies under real-world operating conditions that take into account the complex interplay of AD/ADAS functions, sensor integration, and environmental variability. Regulations such as EU 2022/1426 or NCAP 2029 emphasize the need for evidence-based scenarios and testing under diverse and realistic conditions.
Successfully addressing these challenges is essential to providing scalable, cost-effective, and safe solutions in the automotive sector.
IAV Mela enables users to solve complex tasks with precision and offers solutions not only for passenger cars but also for off-highway applications, including agricultural and construction machinery, heavy-duty trucks, and defense applications. In addition, IAV Mela supports special requirements such as vehicle dynamics and chassis function test scenarios, ensuring robust performance in a wide range of use cases for the overall system.
How does IAV Mela work?
IAV Mela processes textual, either logical or concrete scenario descriptions, e.g., from test specifications, as well as video clips from real-world tests. Using advanced LLM and Vision LLM technologies, it analyzes and categorizes scene elements, extracts the relevant parameters, and thus generates scenario previews and realistic virtual simulations.
Scenarios can be derived directly from requirements, e.g., using IAV devpanion, video data, or sketches, and converted directly into simulation model scenarios. This optimized approach enables efficient testing and development of up to 80%.
Key Features
Automatic creation of simulation scenarios from specifications and textual scenario descriptions
In combination with classic computer vision algorithms, IAV Mela’s Vision-LLM identifies and interprets elements such as roads, traffic signs, and road users, as well as the criticality of scenarios, thus providing a detailed understanding of driving situations.
Use of interfaces to simulation tools such as dSPACE ModelDesk/Aurelion, IPG CarMaker, and ASAM OpenX Standard to convert real-world scenarios into virtual simulations. This enables advanced testing and development in a virtual environment.
Easily create numerous scenarios for comprehensive system testing and validation
By simulating real-world scenarios, IAV Mela supports the development of safer and more efficient driving technologies and paves the way for smarter transportation solutions.
By reducing the need for physical testing and enabling virtual simulations, IAV Mela significantly lowers development costs, making it a cost-effective solution, even for the creation of simulation model tests.
Although IAV Mela is primarily geared toward traffic scenarios, its capabilities extend far beyond ADAS/AD contexts, offering innovative solutions for various areas such as vehicle dynamics optimization, defense systems, mobile machinery and equipment, and much more.
IAV Mela integrates seamlessly with devpanion, Valdivia, and other AI tools from IAV to complete the workflow from requirements to test results and close the loop for continuous integration environments.
Kevin Schössow
