Leading modular automated driving technology supplier aiMotive Kft earlier this month announced the latest release of its aiDrive, which it calls the most scalable, portable, high-performance, hardware-neutral SAE Level 2 to Level 4 software stack for AD (automated driving) and ADAS (advanced driving assistance system) applications. The latest version, aiDrive 3.0, features the company’s virtual sensor technology, which it says enables rapid adaptation to a range of sensor configurations while requiring as little as a fifth of the processing power of conventional approaches.

“Reusability and state-of-the-art perception performance are all essential for the successful and rapid deployment of automated driving in production vehicles, especially for newer players aiming to disrupt the market,” said László Kishonti, CEO and Founder of aiMotive. “The new features in aiDrive 3.0 have already helped us to deliver impressive results for one of our customers targeting SOP in 2024. We believe that by focusing on such key metrics, alongside scalability and hardware neutrality, we have enabled aiDrive 3.0 to set a new benchmark for the automotive industry.”

In addition to upgrades to almost every module, aiDrive 3.0 incorporates many new features.

Among them is VS (virtual sensor) technology, enabling all software modules to be reconfigured to work with a range of different vehicle form factors, sensor types, and configurations with minimal re-validation.

The MS²N (multi-sensor model-space network) approach delivers new levels of perception accuracy with substantial reductions in processor performance (and ECU power consumption) thanks to decreased complexity. By directly fusing data from multiple sensors, pre- and post-processing are greatly simplified, with the resulting model significantly closer to the ideal compared to other conventional ML-based approaches.

DDE (dynamic data-driven evolution) leverages aiMotive’s CI/CD and in-house dataflow management and validation systems to enable continuous enhancements to all software components and features.

The VIF (vehicle intelligence-first) architecture approach means that aiDrive does not need to rely on HD maps for a range of applications.

“Our growing arsenal of unique in-house technologies, complemented by our large-scale aiSim simulation capability and our own automated annotation and continuous fleet testing on three continents, enables us to empower our customers in new and exciting ways,” said Péter Kovács, Senior Vice President, aiDrive.

aiMotive says its aiDrive partner ecosystem has always been central to its product development, as it pursues its mission to bring affordable automated driving to all.

An example is a recent announcement that aiDrive software has been fully ported to BlackBerry’s QNX Neutrino RTOS (real-time operating system), offering safety-critical solutions to partners. According to aiMotive, one of the reasons behind the success of BlackBerry QNX is that its solutions are pre-certified to ASIL D, the highest safety integrity level of ISO 26262. As a result, developers don’t have to go through the hassle of certifying the foundation of their solution, drastically accelerating development times.

“The fact that BlackBerry’s QNX Neutrino RTOS is POSIX-based and looks and feels like Linux, our team felt at home working with the platform from the first minutes,” said Krisztián Holman, Senior Software Engineer of our Embedded Vision Solutions Team. “This allowed us to port aiDrive to the platform easily, using the same framework as we do in our mainline development.”

The AUTOSAR support of BlackBerry QNX enables automotive industry partners to collaborate on larger projects and deliver solutions for mass production.

“It’s great to see aiMotive bring their aiDrive software stack to our QNX Neutrino RTOS,” said Grant Courville, Vice President, Products and Strategy at BlackBerry QNX. “This adds to the growing BlackBerry QNX ecosystem and will enable our customers to accelerate their ADAS and automated drive projects.”

aiMotive also partners on hardware platforms featuring SoCs (systems on chips) from Nvidia, Texas Instruments, Renesas, Mediatek, and Nextchip.

One example of this was a November announcement that Nextchip had licensed aiMotive’s latest-generation aiWare4 hardware intellectual property to deliver convolutional neural network acceleration for its next-generation Apache6. The SoC is designed for centralized domain processors targeting automated valet parking and other demanding automotive processing applications. It combines advanced processors with a rich set of interfaces to enable easy integration into a range of OEM and Tier 1 hardware platforms while offering an elegant software/hardware upgrade path from their Apache5 imaging edge processor.

“Our engineering teams have developed a close working relationship, enabling us to fully benefit not only from the aiWare hardware IP but the broader AI software, system, and algorithm expertise of aiMotive,” said Young Jun Yoo, CMO at Nextchip. “Our SoCs have been focusing on delivering reliable, compact, and affordable solutions. With aiMotive’s expertise, we can make our SoCs more powerful while also optimizing our customer-friendly AI-development environment.”

aiMotive is a privately held company headquartered in Budapest with offices in Munich, Germany; Mountain View and Detroit in the U.S.; and Yokohama, Japan. It grew out of the founder’s first company, Kishonti Kft, at a time of slowdown in the mobile market, developing artificial intelligence and later technologies to support automated driving in 2015.

Kishonti and his top leadership team—Arnaud Lagandré, Chief Commercial Officer; Krisztina Förhécz, Chief Financial Officer; Lajos Németh, Chief Operations Officer; and Dávid Kiss, Chief Innovation Officer—have made aiMotive one of the world’s largest independent automotive technology companies working on Level 2 to 5 automated driving solutions. Its portfolio of software, tools, and hardware products, complemented by proprietary data management tools, enables customers to rapidly develop and deploy production automated driving features. Its lightweight execution stack and sensor-agnostic, reusable data pipeline enable substantial reductions in development costs and timescales.