Mercedes-Benz have partnered with NVIDIA (in a non-exclusive agreement), that will see the pair work together to create autonomous vehicles.
The companies plan to develop the most sophisticated and advanced computing architecture ever deployed in an automobile. There’s no other way to see this as an attempt to come after Tesla, the current autonomous leader.
Nvidia has been developing the Drive platform for years now and this seems like the first serious adoption from a legacy automaker. The biggest problem with the new announcement, is that this technology won’t be available until the end of 2024, or 4.5 years from now.
Mercedes-Benz say that starting in 2024, this will be rolled out across the fleet of next-generation Mercedes-Benz vehicles, enabling them with upgradable automated driving functions. That’s a long time to wait in an industry that’s evolving rapidly.
The challenge of creating an autonomous vehicle is not an easy one. You have to start by establishing the array of sensors, cameras and radar you will select to enable the car to build a 360 degree picture of the world around it. You then need to calculate how many times per second you need to sample that combined dataset and understand the compute requirements of such a task, likely with redundancy (so 2x that number).
The next step is to write the software that takes the raw inputs from the hardware and do two things with it. Respond by controlling the vehicle and use the data to predict what’s is about to happen in the future.
The most common task you think of with autonomous vehicles is to stay within the white lines. In 2020, there’s plenty of systems that are capable of that, but it’s the edge cases that are the difficult part.
A truly autonomous vehicle needs to be able to navigate the dynamic and everchanging world around it, even when there are no white lines. It also needs to keep both passengers and other road users safe. It needs to understand and respond like a human would to the road rules of every different country. The number of challenges to overcome when building an autonomous system is enormous.
Due to how rapidly the world changes, with new roads, whole new estates, roadworks, vehicle crashes, the approach of using high definition maps is just not going to cut it. Those number of variables are so massive that it’s not possible to write an instruction set for each, which means you need to leverage AI to get the job done.
Nvidia are not only working on the chips for the car, but also the server infrastructure to run massive quantities of AI training. Take something like navigating a corner. While your car may have never ever driven around a specific corner, other cars in the fleet, may have driven around corners so similar that the car already knows the appropriate way to respond. To slow to a certainly speed to navigate the corner safely, while also navigating the available drivable space.
When you take this example and apply it to the millions of other decisions that need to be made when navigating a vehicle through this complex world of ours, it quickly becomes a question of how good is your dataset and how much processing power do you have.
Mercedes and Nvidia say that the platform will enable the cars to get smarter over time. Understandably, cars of the future will be always connected.
Tesla pioneered the model of vehicles capturing data from the world and send it back to help with training the AI. Once improvements are made, the whole fleet benefit from the learning via over-the-air software updates.
With so much effort and money required to chase this problem, it’s only thanks to the potential benefits being so enormous that multiple companies are fighting to get there first.
Vehicles that can drive themselves have a massive potential to save lives, enable passengers to be more entertained or productive while in the car. Autonomous vehicles also provide mobility to disabled or the elderly and will change the economics of ride share businesses by deleting the cost of a driver and increasing the available seats in each vehicle by 1.
“We are excited to work with Mercedes-Benz. It’s the perfect partner for us given its long record of innovation and our strong technical relationship. It’s clear from our extensive discussions with Ola and his team that we share a common vision of the automobile of the future.
Together, we’re going to revolutionize the car ownership experience, making the vehicle software programmable and continuously upgradeable via over-the-air updates. Every future Mercedes-Benz with the NVIDIA DRIVE system will come with a team of expert AI and software engineers continuously developing, refining and enhancing the car over its lifetime.”Jensen Huang, founder and CEO of NVIDIA
The new software-defined architecture will be built on the NVIDIA DRIVE platform and will be standard in Mercedes-Benz’s next-generation fleet, enabling state-of-the-art automated driving functionalities. A primary feature will be the ability to automate driving of regular routes from address to address.
In addition, there will be numerous future safety and convenience applications. Customers will be able to purchase and add capabilities, software applications and subscription services through over-the-air software updates during the life of the car.
“We are delighted to be able to extend our cooperation with NVIDIA. Jensen and I know one another well and we have spent a great deal of time talking about the goals and potential of next-generation vehicle computing architecture.
This new platform will become an efficient, centralized and software-defined system in our future Mercedes-Benz vehicles. NVIDIA’s AI computing architecture will help us streamline our journey towards autonomous driving.
These new capabilities and upgrades will be downloaded from the cloud, improving safety, increasing value and extending the joy of ownership for all Mercedes-Benz customers.”Ola Källenius, Chairman of the Board of Management of Daimler AG and Head of Mercedes-Benz AG
A New Computing Architecture
Automated driving functions in future Mercedes-Benz cars will be powered by the next-generation NVIDIA DRIVE platform. The computer system-on-a-chip, called NVIDIA Orin, is based on the recently announced NVIDIA Ampere supercomputing architecture.
The NVIDIA DRIVE platform includes a full system software stack designed for automated driving AI applications. NVIDIA and Mercedes-Benz will jointly develop the AI and automated vehicle applications that include SAE level 2 and 3, as well as automated parking functions (up to level 4).
The new system’s state-of-the-art capabilities will be harnessed with a total focus on safety. As the technology and regulatory framework develop, it will be possible for every car to be updated over the air to enable new automated driving functions. Convenience as well as other safety features will also be available.
To develop the new models, both companies will be utilizing NVIDIA DRIVE Infrastructure solutions to enable data-driven development and deep neural network development to handle the requirements of the regions and operational domains where the cars will be available.
You can watch the press conference for the announced here.