New autonomous vehicle trials seem to be popping up every day. Visit Mountain View, California and you can’t drive a city block without seeing some type of autonomous vehicle navigating the tree-lined streets. We’re not just talking the big letter tech companies – applications developers, technology providers, auto OEMs and every other potential member of the automotive ecosystem is hitting the road to address their own part of the autonomous driving puzzle.
What is the one thing all these players need to pull the future of automotive in closer? A superfast network backbone to enable higher levels of autonomous driving. In any system – biological or technical – the backbone is the fundamental support system which unifies and synchronizes critical system functions.
Just this month, NVIDIA was able to complete a 50 mile, driverless, completely autonomous loop in nerve-racking Bay Area freeway traffic. BB8, the NVIDIA test vehicle, was using the DRIVE AGX Pegasus platform which uses Aquantia AQcelerate 10Gbps Ethernet controllers and PHYs to provide the networking backbone.
In a typical backbone application for In-Vehicle-Networks (IVNs), the 10Gbps Ethernet links can be used for connectivity between: SoC to SoC, SoC to telematics, SoC to storage, as well as the data aggregation links between switches.
The major part of the aggregated speed in the backbone comes from the camera and sensors of the car. Cameras resolutions, which used to be 480p and 720p, are now upgrading to 1080p, and 4k. The pixel size (color depth) has also increased from 8 bits to 16, 20, and even 24 bits per pixel, at a refresh rate of up to 60 frames per second.
The result is that the required data bandwidth to carry high-resolution video is growing from a few hundreds of megabits per second to a few gigabits per second. In addition, the use of raw data is a change for many high-bandwidth automotive sensors. In general, today’s existing high-bandwidth sensor modules such as cameras, lidar, etc., integrate a pre-processing IC to reduce the data bandwidth output.
Level 4/5 autonomy demands a whole new view of the environment, requiring the raw data from various types of sensors to be brought to a centralized processing unit for proper decision-making, for sensor fusion. This approach requires the transport of data from these sensors to the ECUs over the backbone at Multi-Gig rates.
For Level 4/5 vehicle systems, to achieve the goal of avoiding any major failures, the system needs to have enough redundancy for all the mission-critical systems – both for the components and for the network. This backbone network, with the use of switches, is responsible to provide this path-redundancy between multiple ECUs.
Wireless gateways and telematics modules are responsible for the wireless traffic, inside and outside of the vehicle. The telematics module that connects the car to other cars (V2V) and to the cloud (V2I) are required to support increased data rates. These systems are also responsible for the inside-vehicle wireless traffic like Wi-Fi and Bluetooth. In next generation systems, the total generated traffic from the telematics modules is expected to increase to a few multi-gigabits per second.
Data storage on the vehicle is another system that will require a high-bandwidth network. Storage is used for mapping applications – collecting real time road data and images that will be uploaded later to the cloud – as well as to a saved video buffer of a 360-degree-view of the vehicle, in case of an accident which could be later used in court if necessary. OEMs can use the storage for other specific applications. Not all traffic to the storage requires ultra-high-speed bandwidth, but due to the bursty nature of some applications, and the need for others for continues high-bandwidth data, the data link to the storage need to be at a speed of multi gigabits per second.
With the growing number of high-bandwidth systems in the vehicle, as well as the number of Multi-Gig camera and sensors, the backbone, which aggregates and switches all this traffic, needs to be upgraded to automotive Ethernet links at rate of 10Gbit/s which are enabled by products like Aquantia’s AQcelerate controllers and PHYs. The NVIDIA autonomous loop is an excellent proof point for why a 10Gbps Ethernet backbone with AQcelerate products is essential to the successful delivery of safe, secure autonomous vehicles.