The one thing that Cerence + Smart Eye + Mercedes all have in common is NVIDIA DRIVE IX.
View attachment 32430
- Nvidia worked with Mercedes-Benz on MBUX Hyperscreen.
- And there are the vehicles running Nvidia-powered Cerence Look using eye-tracking from Smart Eye, such as the Mercedes-Benz EQS.
- #25,287
- The article (link) below states:
- "Using NVIDIA GPU technology, Smart Eye has been able to speed up its cabin-monitoring system — which consists of 10 deep neural networks running in parallel — by more than 10x"
- "In-vehicle technology companies Cerence, Smart Eye, Rightware and DSP Concepts are now using the platform to deliver intelligent features for every vehicle occupant."
Confused? So am I, but there's a reason why it has been suggested that NVDIA are more like a partner to us, than a competitor.
Inside AI: NVIDIA DRIVE Ecosystem Creates Pioneering In-Cabin Features With NVIDIA DRIVE IX
Cerence, Smart Eye, Rightware and DSP Concepts are now using the DRIVE IX platform to deliver intelligent features for every passenger.blogs.nvidia.com
Someone posted this yesterday:
Mercedes MB.OS
Designing our own chip-to-cloud architecture as a car manufacturer? Yes, please!
Magnus Östberg Chief Software Officer at Mercedes-Benz AG March 15, 2023
(18) Designing our own chip-to-cloud architecture as a car manufacturer? Yes, please! | LinkedIn
The link says:
...
We maintain three chip categories, including custom high-performance system-on-chip (SoC), mid-level MB.OS base layer microcontrollers, and low-spec chips supporting industry standards. This enables independent strategies at the chip level, including our custom automated driving technology, enhanced UI/UX and wireless connectivity, all the way down to standardized interconnectivity, like USB controllers.
...
Fast software is enabled by our state-of-the-art SoC microcontrollers, which target a high ratio of performance to efficiency. Through standardization of important infrastructure components we can cover more than 90% of our MB.OS architecture with chips from a small number of suppliers, with tailored strategies for each based on future application requirements and learnings from the chip crisis.
Our keys to success at scale will include direct, long-term sourcing and increased in-house development
parts of MB.OS, like our Infotainment system MBUX. However, we also collaborate with partners to employ hyperscaler solutions including Microsoft Azure and Amazon AWS, complemented by partnerships with Google GCP in western regions and Tencent in China.
…
A core feature of MB.OS is lifetime learning powered by data. As our systems become even more data driven, they will be able to learn on the road, taking advantage of our best-in-class sensor setup across the entire customer fleet. Our vehicles can collect up to 300 petabytes of data every year to further train our neural networks.
Our advanced in-vehicle machine learning algorithms automatically chooses the data that is most useful for improving our software stack, whether it is creating or updating an HD map or improving our perception algorithms. This means our next-generation automated driving system will continue to improve throughout its life in customer hands.
We are working on new driving and parking functions that can understand and manage infrastructure changes, including new traffic signals or road layouts. They will be able to adapt to diverse geographies around the world, and continually improve their performance via AI to learn safer, human-like driving behavior.
Partnering with industry leader NVIDIA to develop a revolutionary in-vehicle computing and software architecture enables advanced autonomous driving, with complete computing and AI infrastructure to support it. We’re building this full-stack platform jointly, including the operating system, middleware, and all the algorithms to enable advanced autonomy in all sorts of different scenarios.
Mercedes have 3 flavours of chips:
custom high-performance system-on-chip (SoC);
mid-level MB.OS base layer microcontrollers; and
low-spec chips supporting industry standards.
Mercedes are standardizing their component inventory:
"Through standardization of important infrastructure components we can cover more than 90% of our MB.OS architecture with chips from a small number of suppliers"
When your on a good NN, stick with it!
Mercedes are partnering with Nvidia:
"Partnering with industry leader NVIDIA to develop a revolutionary in-vehicle computing and software architecture enables advanced autonomous driving, with complete computing and AI infrastructure to support it."
Is Nvidia our friend?
Does Nvidia have a NN friend?
We are friends with Valeo.
Valeo is doing the Mercedes Level 3 LiDaR, but they are being supplanted by Luminar for the next round.
Luminar and NN? Not so squeezy.
Will Mercedes have to throw out a warehouse full of Valeo/Akida LiDaRs?