Mobileye consumer AV push: facts behind the math?

By Junko Yoshida for the Ojo & Yoshida Report – Mobileye insists consumer and industrial AV development must proceed in tandem if both sectors are to succeed.

The AV industry has generally come to accept that the highly automated vehicle market will start with robotaxis, scheduled for rollout one city at a time. Consumer-owned AVs without human drivers are believed to be years — some say decades — away. But Mobileye, whose parent company Intel plans to take it public later this year, is pushing an aggressive timetable for consumer AVs, warning the industry that consumer and industrial AV development must proceed in tandem if both sectors are to succeed. For now, OEMs are left with the daunting task of checking Mobileye’s math.

While others in the AV industry have focused on robotaxis, Intel subsidiary Mobileye has stood out by doggedly — and publicly — attacking the seemingly intractable problems that stand in the way of building fully autonomous vehicles consumers can buy. At CES 2022 this month, Mobileye doubled down on its prediction that consumer AVs capable of driving everywhere will emerge in the 2024–2025 timeframe at an added cost of less than $5,000.

This position, confidently argued at CES by Mobileye CEO Amnon Shashua, makes the future of consumer AVs seem plausible, except to the ranks of diehard AV skeptics.

Mobileye believes it has math, data, and experience on its side. The company’s holistic measures range from the development of HD mapping, driving policy, and AV/robotaxi trials in various cities to the advent of lean-compute AV systems-on-chip, tightly coupled software and hardware, and true redundancy in sensing.

Do those advances put Mobileye in a league of its own? “I wouldn’t go that far, but they are certainly unique,” Phil Amsrud, senior principal analyst, automotive, at IHS Markit, told the Ojo-Yoshida report. Given Mobileye’s vertically integrated approach to AV development, Amsrud acknowledged that Mobileye is no longer just a regular chip supplier. “I see them as Tier 1.5,” he said.

Consumer AVs are the cornerstone of Mobileye’s vehicle development strategy. While confirming that Mobileye is pursuing parallel development of robotaxis and consumer AVs, Shashua stressed that this is “not just about hedging.”

Noting the “very strong synergies between the robotaxi and the consumer AV,” Shashua argued that if a company starts AV development with only robotaxis in mind, scaling the technology down to consumer AVs will prove difficult. In contrast, developing a purpose-built consumer AV creates technology that cuts the cost for robotaxis, enabling scalability in that business.

Of course, some industry observers staunchly insist that the era of consumer AVs will not arrive for decades. In Shashua’s view, however, forestalling the development of consumer AVs would be a mistake. The AV industry can build robotaxis loaded with compute power and all the sensors money can buy. But without pursuing fundamental innovation from the get-go in architecture and technologies to make consumer AVs safe, reliable, and affordable, OEMs might never move past the industrial AV market.

If Shashua is right, lingering consumer AV hesitancy could hamper the successful launch of an economically viable robotaxi business.

Industry analysts said they were not surprised by Mobileye’s consumer AV push. But they differ on why the company is pressing so hard.

Tom Hackenberg, principal analyst for computing and software at Yole Développement, said the consumer AV development goal forces the company to set a high bar for iterative AV development. “They can learn what worked and what didn’t work.”

Hackenberg also suspects that the push for consumer AVs is likely coming from Mobileye customers who don’t want to be sidelined by companies like Tesla down the road. Nonetheless, a 2024–2025 timeline for rollout of consumer AVs with full operational design domains (ODDs) “might be a little optimistic,” he said.

IHS Markit’s Amsrud said Shashua believes the consumer AV is “the holy grail. If they can make self-driving cars that you and I can buy, Shashua believes that Mobileye will have really built something that progressed the industry.”

Amsrud added, “If I want to be a volume producer, robotaxis are a stepping stone, but not the final destination. There will be a lot more consumers buying cars than robotaxis being produced.” 

Many AV industry sources say the key difference in capabilities between Level 2 vehicles and Level 4 driverless vehicles lies in the ODD — that is, where and under what conditions the vehicle can drive safely.

Shashua disagrees — and not surprisingly, given Mobileye’s plans to support full ODDs starting with Level 2.  “The differentiating factor between Level 2 and Level 4 is not the ODD, but the mean time between failures,” he said.  If the MTBF is lower than the average human driver’s statistics, it’s L2. If the MTBF is higher, “it can be an L4 system.”

In his opinion, highly automated vehicles “should have a very, very high mean time between failures — much higher than the average human driver, somewhere [from] 10 times higher [to] 100 times higher, 1,000 times higher.”

Other than Mobileye, no carmakers, sensor companies or automotive chip designers have been willing to discuss MTBF goals.

Meanwhile, consumers generally remain skeptical of the safety and reliability of AVs.

The AV industry’s original sin was to encourage expectations that “AVs can make all the human drivers’ driving errors go away,” said Amsrud. This isn’t true; in fact, the more trials done, the more problems — many completely unforeseen — have popped up.

“It just worries me that while the technologies going into AVs are so sophisticated, there are still going to be issues,” he said… Full article

Headline image: Mobileye CEO Amnon Shashua at CES 2022 – Courtesy of MobileEye, 2022