What the hell is Amazon doing in the auto business? The answer is simple: AWS is committed to developing networked vehicles.
In November 2020, NXP Semiconductors entered into a partnership with AWS. The purpose of this deal is to enable automakers to collect and use the vast amounts of data generated by their cars.
The automotive industry has long been talking about connected vehicles. Communication modules installed in cars have allowed companies to create telematic services such as General Motors' OnStar. These technologies also allow customers to download applications and other content for infotainment systems.
“There are phases one and two in connected car development,” said Brian Carlson, director of global marketing at NXP Vehicle Management and Networks. In the third phase, the partnership between NXP and Amazon will allow “automakers to collect data from all vehicles,” he explained.
But what data are we talking about?
With a high-performance, high-bandwidth vehicle network processor (such as NXP's S32G), carmakers will be able to push data to the cloud, "from sensor readings to algorithmic and behavioral statistics," Carlson said.
Some data will be processed using edge computing, and some will be directly in the car. The NXP-AWS deal will give OEMs the ability to mine and explore (in the cloud) data that hasn't previously been analyzed.
In a world of Internet of Things where connectivity is critical, the engineering community has moved from personal desktops to the cloud using services such as AWS and Microsoft Azure.
Connected cars are next. “This will be a new trend,” said Egil Juliussen, a seasoned analyst in the automotive industry. Much of the work of developers and engineers goes to the cloud because “cloud services provide a lot of advanced tools,” he said.
How exactly will engineers use cloud services?
First of all, they will focus on improving the perception of self-driving systems and ADAS, as well as transferring machine learning algorithms to the cloud. Ideally, the safety of car networks and the health of electric vehicles' batteries can be monitored from the cloud using real-time vehicle data.
AI in the cloud
Many automotive features are already being developed using the cloud. For example, in the cloud, you can design, train, optimize, and deploy machine learning models for cars.
What additional benefits could NXP and Amazon bring together as AI development using the cloud is commonplace? Carlson believes the use of real data is key - it can help developers improve performance and security. He also added that real-time communication allows automakers to detect and record various critical cases and anomalies.
So, which AI processors will NXP and Amazon be working with?
As it turns out, AWS SageMaker Neo can work with different machine learning engines and is even optimized for SIMD devices. "This list includes x86, Arm, RISC-V and other architectures," Carlson explained. It will also use the PCI-Express interface built into NXP's S32G to provide support for a variety of processors ("from Nvidia to FPGAs and Snapdragon") for autonomous driving systems and ADAS. TPU from Google will soon be included in the list of supported devices. which are not currently supported.
Carlson also said that it is very important for an interface chip manufacturer to make their products flexible and processor independent. "Our products must provide the ability to embed AI in cars," regardless of whose processor they use.
Also, let's not forget that NXP has solutions for machine learning - in particular, the S32V machine vision processor. NXP has developed the eIQ Auto toolkit that will accelerate the "quantize, shrink, and shrink" neural networks. Much of this is done in the cloud using data from cars.
Service-oriented interfaces
Through its expertise in handling automotive data and leveraging AWS cloud infrastructure, NXP has introduced a number of new services that automakers can implement for themselves.
Sending data generated by lidars, cameras and other sensors to the cloud for analysis, as noted above, is an important first step towards improving sensing systems in self-driving vehicles and ADAS.
As for NXP, Carlson noted that "EV development" is the main scenario in which the collaboration between NXP and Amazon will be visible. The Edge Computing Cloud Model allows real-time monitoring of the health of batteries, motors, and other components. According to him, working with digital "twins" of real cars in the cloud will optimize energy consumption and expand the range of electric vehicles.
Security is another area where data from connected cars will play an important role. “Think about network intrusion detection,” said Carlson. “By pushing data to the cloud, machine learning models can improve security and updates can be deployed to entire fleets to prevent breaches.”
Many automakers also expect networking features to simplify “vehicle health management”. Carlson explained that real-time monitoring of automotive data using edge computing, combined with machine learning, will enable automakers to find problems "before the car even knows about them and issues an error code or the check engine fires."
Also, a very important component of support for highly automated connected vehicles will be the analysis of the operation of driver monitoring systems in real time.
Changing the role of network processors
Expanding access to automotive data is critical for automakers willing to implement OTA updates into their solutions.
Equally important are the characteristics of network processors (such as the S32G) and their interoperability with various electronic control units (although many of these devices may be supplied by third parties). “From a network perspective, the goal of NXP is to provide OTA updates for all control units in the system,” Carlson said.
Over time, as telematics systems proliferated, automakers also began distributing software updates over the network. However, key updates and maintenance are done through the OBDII ports. The release of more powerful network processors will allow automakers to develop their cloud services, Juliussen said. This is an area that many want to get involved in because it could potentially be very profitable.
In theory, major automakers (such as GM and Toyota) can build their own cloud platforms. As Juliussen said, this is what they are doing.
Carlson added that these companies are also integrating ready-made cloud services such as AWS and Microsoft Azure into their solutions. The point is that tools from Amazon and Microsoft can make it easier for automakers to develop software and various services.
According to NXP, the S32G is significantly different from many other network control units used in the industry. The company also argues that the role of traditional networked control units is limited to the safe movement of data within the vehicle. Carlson noted that "the S32G, in turn, can accelerate network processing, run high-performance real-time applications and transfer data securely to the cloud."
Now that manufacturers are looking for more and more new tasks for the S32G, Carlson jokingly said that “now we say that the G in the S32G means that it is a general-purpose processor (from the word general - translator's note)”.
As a reminder, the S32G is an ASIL D-compliant automotive network processor that provides hardware security, high real-time performance, and a variety of applications. and accelerating networking for service gateways, security controllers and processors. "
The processor uses Quad Arm Cotex-A53 cores (using Arm Neon technology), organized into two clusters of two cores with cluster blocking for individual applications and services. It also features Lockstep Triple Arm Cortex-M7 cores for real-time applications, a low latency networking module and a packet forwarding engine to accelerate Ethernet.
As powerful as the S32G may seem, Carlson stressed that the purpose of the NXP gateway processor is not to send terabytes of raw data to the cloud.
NXP partners with Teraki and SafeRide to reduce the amount of data sent to the cloud. To do this, companies look for anomalies and apply analytical algorithms. Noting that the company is aware that it is looking for a needle in a haystack, Carlson explained that the main purpose of the network processor is to transmit only what is important.
The conclusion is that the development of new cloud services requires the cooperation of different companies. “Right now, we at NXP are developing partnerships,” said Carlson. He added that the collaboration between NXP and AWS is just the beginning in terms of what can be gained from the new infrastructure.
- Russia's first serial control system for a dual-fuel engine with functional separation of controllers
- In a modern car, there are more lines of code than ...
- - Automotive, Aerospace, (50+)
- McKinsey: automotive
