Razvan Has, a staff systems engineer at PlusAI, which released SuperDrive 6.0, the latest version of its autonomous driving software engineered for commercial-scale deployment earlier this year, sat down with the AAVI team to discuss scalable V&V frameworks for Level 4 autonomous trucking during the conference at Autonomous Vehicle Tech Expo.
Has’s presentation on Day 2 described PlusAI’s systems and safety engineering framework, which Has said has been explicitly tailored to overcome the “validation mileage fallacy” for Level 4 autonomous trucking deployments globally. He explained how the company has decoupled execution spaces into distinct problem, solution, verification and validation layers to ensure that engineering overhead converges toward zero as new commercial missions are expanded.
Has went on to share how to systematically orchestrate scenario-driven decomposition, event replay and multi-tier software-in-the-loop (SIL) and hardware-in-the-loop (HIL) test environments. Delegates were also able to discover how PlusAI uses machine learning-style transfer to synthetically expand real-world datasets across challenging operational design domain (ODD) conditions like rain, fog, night and snow; and how it aligns non-negotiable safety cases with technical compliance and commercial business goals.
“At PlusAI, I am responsible for developing a set of methods that can be applied to create a concrete framework on how best to execute and engineer a digital driver for a very specific mission,” Has told AAVI after his presentation. “[As a hub-to-hub autonomous trucking partner] you can think of us as connecting two dots on the map, which allows us to really focus our work. My presentation laid out a framework that anybody could theoretically pick up and execute to help them scale. It’s very cool to interact with such a concentrated population of experts in autonomous driving, where you can really get to know each other over the three days of the event.”
