Mercedes-Benz Partners with NVIDIA to Enable Its Future Software-Defined Vehicle Architecture

In June 2020, Mercedes Benz announced a landmark partnership with NVIDIA to develop a new, software-defined vehicle architecture, set to be introduced in 2024 before being rolled out across the entire Mercedes-Benz vehicle lineup. First announced in December 2019 (following earlier teasers), NVIDIA’s ORIN System on Chip (SoC) will form the basis of the new Mercedes-Benz Autonomous Vehicle (AV) architecture, with its impressive 200 TOPS performance providing the necessary headroom for new AV features to be deployed Over-the-Air (OTA) throughout the vehicle’s lifecycle. This represents a significant departure from legacy automotive architectures, not only in terms of raw performance, but also in terms of design philosophy, with a renewed focus on post-sale revenue opportunities. The partnership sheds further light on how to scale the engineering effort involved in transforming the automotive vertical so that it can align with the consumer expectations set by consumer electronics, particularly in regard to mission-critical autonomous driving functions.

Where Tesla leads, the rest of the automotive industry follows. This was certainly true of electrification and has now filtered through to attitudes toward software upgrades and lifecycle maintenance, even in the mission-critical domain of autonomous driving. Tesla’s market leadership notwithstanding, the whole automotive vertical is, in reality, learning to imitate the same development philosophy and design strategies pioneered by other software-defined consumer devices.

Legacy automotive thinking sees new features and functionalities as differentiators that should be leveraged to drive the shipment volumes of new models. This has led Original Equipment Manufacturers (OEMs) to opt for the minimum hardware specification possible to enable a desired function/feature, allowing for a certain model to be differentiated, at minimal cost, in order to drive sales of that model. The new OTA paradigm will place priority on “headroom,” shipping vehicles with more hardware capacity than necessary at point of sale to allow for meaningful OTA software updates over the course of each vehicle’s lifecycle. With an aim to eventually equip every model with NVIDIA’s 200 TOPS ORIN SoC, Mercedes-Benz will, over the course of the next decade, gradually build up an installed base of vehicles with the capacity for OTA delivery of meaningful autonomous functions and therefore considerable scope for monetization post sale.

The NVIDIA/Mercedes announcement heralds the arrival of the OTA paradigm into the automotive mainstream, and further confirms the trend for connected cars to be designed and monetized in the same way as other connected devices, such as smartphones, tablets, and smart home devices. When pursuing a strategy so radically different from the status quo, there are approaches that OEMs can take to minimize risks and maximize the OTA opportunity:

• A Clear Roadmap, with Multiple L2+ Applications: The greatest risk facing an OEM when adopting the OTA paradigm is the failure to take advantage of the extra capacity included at point of sale at extra cost. For example, Audi launched a modest L3 application in the form of its Artificial Intelligence (AI) Traffic Pilot, a system enabled by its Central Driver Assistance Controller (zFAS) platform. However, the system ran afoul of AV regulations in key markets, with the result being that there are Audi A8s roaming on the road with hardware capacity far in excess of the L2 applications currently permitted in most regions. Given that predicting the pace of regulatory development over the course of the next 10 years is fraught with difficulties, it is vital that OEMs have a strategy to develop and deploy autonomous functions within the context of L2+, in order to provide some certainty of Return on Investment (ROI).
• No Half Measures: Unlike the smartphone market that OEMs are looking to imitate, passenger cars have a long development cycle and an even longer lifecycle on the road. It’s important for OEMs to correctly anticipate the hardware requirements of the autonomous functions they wish to deploy over the course of 5 to 10 years, and resist the pressure to minimize the hardware specification to what is required at point of sale, in order to cut production costs. The 200 TOPS of processing power in NVIDIA’s ORIN SoC offer considerable headroom above the current crop of autonomous applications, with support for applications up to SAE Level 4. Anticipating hardware requirements will not only be a matter of equipping headroom, but also equipping the necessary Human-Machine Interface (HMI) technologies (such as camera-based driver monitoring systems) to make L2+ applications feasible.
• Hardware Modularity: While the hardware requirements for safe autonomous driving are better understood than the software requirements (particularly with respect to validation), a scalable and modular hardware platform can reduce the risk of system irrelevancy as hardware capabilities continue to develop. A core design philosophy of NVIDIA is the consistency and scalability of its hardware architecture, enabling its customers to retain their software development efforts between iterations of the NVIDIA DRIVE platform.

Ultimately, a successful autonomous OTA strategy will require OEMs to have an accurate, long-term understanding of the future hardware requirements of the applications they want to monetize over the coming 5 to 10 years, committing to seed the market with headroom with a view for long-term monetization.