This year, Intel is starting the 2021 trade fair at CES with not just one, but two press conferences. The company reserves its computer announcements for the afternoon from 1:00 p.m. PT. Then we expect more about the Rocket Lake 11th generation desktop processor platform and much more.
In the meantime, Intel plans to use the morning hours to announce the latest advances made by the Mobileye business unit in autonomous driving.
We'll be updating this post throughout the week of CES, but here's everything Intel has announced so far.
Intel's morning CES session is entirely dedicated to autonomous driving through the company's Mobileye unit. The press conference chaired by Mobileyes CEO and Intel SVP Amnon Shashua has just begun. We'll update this post as we get more announcements. If you're expecting PC announcements, be prepared this afternoon for news regarding Intel's processors as we will update this post later in the day.
"Mobileye is a growth engine for the company and Intel is deeply committed to the business," said Shashua of this year's Intel press conference.
Mobileye is expanding its test scope. Shashua talked about driving policy algorithms and making high-resolution maps as factors that affect scalability. The company believes its driving policy is now transferable so it can open up new markets for testing. Now, after five years, the company feels it can create large-scale production maps.
As COVID ravaged the planet, the company said it adapted its business model to scale to open up new markets. Rather than building an entirely new facility, Shashua said Mobileye was able to ship a vehicle – in two weeks with two sales representatives – and demonstrate it to OEM partners in Munich.
"This made us feel like we can scale now," he said. The company plans to expand worldwide to include Shanghai, Tokyo, New York and Detroit.
Mobileye said it measures failure rates in hours of driving. In the United States, a crash occurs approximately every 5,000 miles. This is roughly one in approximately 50,000 driving hours.
With 50,000 cars, this means that on average we have an accident every hour that can be traced back to human driving. With autonomous driving, however, this is not acceptable – it has to be better. For this reason, Mobileye developed a redundancy system to reduce the crash of a Level 4 system.
Cameras, radar, and lidars help create this redundant system for autonomous driving, Shashua said.
"It's so important from a technology standpoint to do the hard work," he said. It's not just about radar and lidar. The company is approaching autonomous driving by first building an end-to-end camera system and then adding radar and lidar as a redundant system.
Mobileye is also building a level 2 system for Chinese markets. Since it is a camera-based system, it would be affordable for consumers, the company said. This is the route Mobileye is taking to monetize its efforts before reaching level four reliability.
The company anticipates autonomous consumer vehicles will take some time to complete, so they want to tackle robo-taxis first. The robo-taxi will be built with luminar lidars and no cameras, but the cameras will be added before takeoff.
In 2025, Mobileye is looking at the next generation of lidars called FMCW. Intel has an advantage in this area, said Shashua, with photonics with active and passive layers on chips.
On the radar side, the company is targeting imaging radar that has higher resolution and is more software defined than analog.
And when the technologies mature, Shashua expects we'll have affordable tier four autonomous vehicles by 2025.
Shashua claimed that its Responsibility Sensitivity Safety (RSS) framework is one of the company's greatest achievements as society will not accept misjudgments with a computer. With autonomous vehicles, it has to behave with human judgment because it interacts with people. And the idea of being careful – giving in when you don't have the right of way – is not mathematically defined. RSS is a mathematical definition of what it means to be “careful” while driving. The framework takes the worst case scenario with the assumptions that have been defined with the regulators. It's a rule-based way of thinking about what is reckless and what is careful, Shashua said. The theory must be completely transparent.
"RSS is one of our crown jewels, and we deliberately made it transparent," said Shashua. Essentially, RSS gave Mobileye a computation of human behavior while driving that goes beyond red lights and speed limits.
When building a perception system, the bottleneck is understanding all the risks and the semantics of the road and not identifying any people or objects on the road. It's about understanding the priorities along the way to determine who has to give in, Shashua said. "It's so detailed that the probability of not making a mistake on a single path is almost unattainable, =,"
By piecing together information from multiple cars, the company realized it could do better than all of the data gathered from a single car. This is known as creating a map – not the navigation map from Google Maps, but data from cars. This is what Mobileye defines as REM technology, or Road Experience Management. The company wants to capture low bandwidth data and do hard work because uploading detailed, high bandwidth data would be costly and OEM partners may not want to invest in that bandwidth. Data collection isn't about recording events, he said. The point here is to limit the complexity.
The company now has nearly 1 million vehicles from six automakers that collect data and send it to the cloud. Worldwide coverage is global and scalable.
Intel completed the acquisition of the company in 2017 with an estimated enterprise value of $ 14.7 billion.
"The acquisition combines the industry-leading technologies of both companies, including Intel's expertise in high-performance computing and connectivity, and Mobileye's leading expertise in computer vision to develop automated driving solutions from the cloud to the network to the car." so the company said the time.
The acquisition has enabled Intel to distinguish itself as a self-driving platform and to expand beyond pure computing. In this area, it has to compete with technologies developed by Google's Waymo and Nvidia Drive platforms.
Intel and Nvidia have been competing more and more against each other in recent years. Intel's entry into the graphics processor sector also encroaches on an area that is dominated by Nvidia's GeForce RTX graphics cards. Recently, and ahead of CES 2021, automaker Nio announced it was partnering with Nvidia on the company's Drive platform to bring advanced autonomous driving capabilities to market.
The computer keynote is expected to start at 1 p.m. PT. You can watch the livestream from the Intel website, but we'll also cover the latest news when we update this post.
In addition to the latest news from Intel, be sure to follow Digital Trends' CES 2021 hub for the latest technical news and announcements from the show, as well as hands-on sessions on products and in-depth analysis.