Mercedes-Benz and NVIDIA have announced that they will cooperate on the development of an in-vehicle computer system and AI computing infrastructure. The new technology is expected to be rolled out across all Mercedes-Benz model lines starting in 2024 to provide next-generation vehicles with upgradeable, autonomous driving functions.
The aim of the collaboration is to develop intelligent and advanced computer architectures for all Mercedes-Benz model series.
The new software-defined architecture is based on Nvidia Drive and, as a new standard, should enable regular routes to be driven automatically. According to the companies, there will also be numerous other safety and comfort applications. Via over-the-air updates and cloud applications, customers will be able to purchase and add new software applications and subscription services throughout the entire life cycle of the vehicle.
The new CEO partners, Jensen Huang and Ola Källenius, have probably discussed their common vision in detailed preliminary talks and want to use a variety of AI and software tools to further develop and improve the car and its life cycle.
The new computing architecture
The automated driving functions of future Mercedes-Benz vehicles will therefore be driven by the Nvidia Drive platform. The System-on-Chip (SoC) Orin is based on the ampere supercomputing architecture recently announced by Nividia. The drive platform includes a complete system software stack specifically designed for automated control of AI applications. Nvidia and Mercedes-Benz plan to jointly develop AI applications and automated functions that include level 2 and 3 levels of autonomous driving and automated parking functions up to level 4.
The focus of the new system will be on safety. With the development of the technology and if the appropriate legal framework is in place, every car should be able to receive new automated driving functions via over-the-air updates. The collaboration is based on Nvidia's drive infrastructure to enable both data-driven development and the use of deep neural networks. The vehicles could then also be adapted to the individual regulations and requirements of countries and for specific deployment scenarios.