General Motors will give a soul to all its new cars (digital shell)

The company is currently developing a new “digital nervous system” that supports auto-update software and processing up to 4.5 TB of data per hour


image

Photo of GM Cadillac CT5 2020 prototype.

New “electronic platform” - a computer network that will run on the vast majority of the company's vehicles and ensure the operation of their many digital systems. It is just as important to the future automaker as any single function or even the car itself. It is this infrastructure that will enable GM to compete in an industry that is increasingly dominated by software products and provide its customers with all the high-tech benefits they need, from high-resolution screens to amazing security features.

“It's the car’s brain and nervous system,” says Al Adams, GM's director of electrical components and subsystems, who led its development. Think of it as an infrastructure that allows all parts of the car to communicate clearly and safely. You could also call it guts - a low-value, hidden part of the body that is not noticed while it is working. Architecture will debut on the Cadillac CT5 2020, due to appear later this year. Over the next four years, GM will introduce it to most of its global lineup, which has 56 models of four brands.

So, more cars will receive a system of semi-autonomous driving Super Cruise Cadillacand other active security features. Now GM will be able to release software updates over the air, improving the performance of its engines or the performance of suspensions on rough roads, even years after the sale of the car. Greater computing power improves screen resolution. More intelligent battery management systems can squeeze more miles from electric vehicle batteries.



“We must completely rethink the computer system”


For GM, the process began a few years ago. According to Adams, this is part of ongoing evolution. Modern cars contain up to 30 or 40 computers that control everything from transmission to a small light that turns on when the user presses the window lock button. This is about twice as much as ten years ago, even though the GM team combined various functions in one computer. They also worked to minimize the number of wires passing through the car and weigh several kilograms.

Scientists from the University of Maryland shared their opinion that the AI ​​method used today is not suitable for full-fledged autonomous vehicles. They proposed to replace classical neural networks with algorithms based on hypervectors. Such AIs will be able to form memory and reflexes.

We also thought about hackers


The concern claims that cybersecurity has been introduced from the very beginning. “Being cybersafe is of utmost importance to GM and is a prerequisite for entering this space. We conducted a comparative analysis with the defense and aviation systems, ”Adams said. And the company actually took a rather proactive approach to cybersecurity: in addition to the internal red team, it also uses third-party testers and the Hacker One community to search for software exploits. Eliminating these exploits has the same priority as any other vehicle safety issue. The new platform includes message authentication: when various modules in the car exchange data, the receiver and transmitter use the “security handshake” used for each piece of protected data.

“I will never buy a car that is always connected to the Internet”


All this is great news, not only for most motorists, but also for software developers. The implementation of centralized control of car systems with Internet access by one of the leading automakers opens up many new fields of activity and provides jobs for young IT specialists. However, not all users are happy with such an initiative. Many car owners are supporters of the retrograde worldview and believe that developers will intentionally leave vulnerabilities for earnings on fixing errors that occur during operation, as is often the case with mobile devices and operating systems. For an in-depth study of public opinion, go to the Reddit news discussion page .

So far, only Tesla AI car steering wheel AI used for their driver assistance system can be represented as analogues - a kind of modernized “cruise control” that can keep the selected lane on the highway, but does not relieve the car owner of the responsibility associated with driving.

To whom this topic is interesting, you can find the following links:

Hackers forced the Tesla autopilot to “meet”
Tesla will focus on weak AI and neural networks in the new generation Autopilot.
The battle of the neural network and lidar
General Motors showed the work of its autopilot in the video

Sources:

Arstechnica
Wired

Also popular now: