Does FCA Have A Secret Self-Driving Car Project?
Reading between the lines of an obvious leak.
Fiat Chrysler has a secret self-driving car project, according to The Information, who make the claim in an extraordinarily vague op-ed entitled “Fiat’s Secret Self-Driving Car Shows How The Industry May Evolve.”
So, does FCA actually have a secret self-driving car project?
I’m a big fan of The Information, but there’s so little information in their article that it’s almost impossible to know what FCA has, if anything. Let’s deconstruct the article paragraph by paragraph and try to separate fact from conjecture.
“Fiat Chrysler has garnered outsized headlines for becoming the first traditional carmaker to make electric minivans with self-driving technology designed by Alphabet’s Waymo.”
Fact. Not a surprise as much as an insurance policy against total irrelevance in an autonomous future. FCA had never previously admitted to the existence of any self-driving research program, so any deal with Waymo opens the door for them to become Waymo’s automotive Foxconn, a role Daimler famously rejected but FCA would seem to need.
“Behind the scenes, however, Fiat Chrysler is developing its own autonomous vehicles so that it won’t be just a metal bender for tech companies like Waymo, The Information has learned.”
Really? This suggests FCA has a real program, as in a team capable building a self-driving car, which means they now join the other thirteen major manufacturers with a similar program with varying degrees of progress.
Or does it?
Where’s The Beef?
The next paragraph is really confusing:
“Details of the previously undisclosed program underscore how automakers are grappling with the shift toward automated vehicles. Both Fiat Chrysler’s two bigger U.S. rivals, Ford and General Motors, are also working on their own self-driving cars.”
True, Ford and GM are working on self-driving cars, but no details that follow suggest FCA’s program is anywhere near those of Ford or GM, whose vast and well-funded programs combine both in-house and external R&D; i.e. Ford’s investments in Velodyne and Civil Maps, and GM’s $1B acquisition of Cruise Automation.
“And like those companies, Fiat Chrysler plans to integrate self-driving components designed by others to fit its unique technical requirements, thereby creating a differentiated riding experience.”
I’m not sure I understand the point being made here, at least not yet.
“And like those companies...FCA...plans to integrate...components made by others…” makes sense, I suppose, because no car maker in their right mind is designing and manufacturing self-driving car hardware from scratch, except maybe Tesla. Virtually everyone is buying the same camera, radar, and ultrasonic hardware from Tier 1 suppliers like Bosch, Continental and Mobileye. FCA is just far behind in integrating them.
What does “unique technical requirements” mean? Other than fitting the same hardware everyone else is using into FCA’s cars, there are no unique technical requirements. And what does “creating a differentiated riding experience” mean? Is The Information referring to an extra software layer a la Tesla Autopilot, which sits on top of Mobileye’s? No way of knowing from their article. I have a guess, but we'll get to that.
We’re only two paragraphs in, and all I can glean was that the someone at FCA, or someone linked to FCA’s publicist, shared some BS laden self-driving car terminology with my friends at The Information.
Let’s keep going.
“That approach reflects the view, held by Fiat Chrysler executives and others in the industry, that the core self-driving technology will become widely available because so many companies are working on it.”
This makes sense all around, but then we have—
“So rather than focus on its own version of that core tech, carmakers will invest in ways of customizing the technology. Among the differences will be how much control over the vehicle the passengers can choose to exercise.”
Huh? Again, no car makers (that we know of) besides Tesla are working on core tech, so this is not new. Everyone is building custom software and interface layers on top of commoditized hardware, with minimal to marginal degrees of success.
As for differences in control over the vehicle, this is not a technical issue as much as a legal and political one. Tesla has aggressively marketed Autopilot by combining their state-of-the-art lane keeping with the longest hand-off intervals of any semi-autonomous suite on the market.
Competing suites are all in orbit around their off-the-shelf hardware, with minor variations in automatic emergency braking (AEB) reaction times and middling lane keeping.
The Information then names the principals behind FCA’s secret self-driving team:
“Key parts of the program are led by Justin Carlson, who has spent about a decade overall at Chrysler, most recently working on “driver assistance” tech such as adaptive cruise control. The company has hired many others, including Miguel Hurtado, a Ph.D. in detecting objects using sensors on a car, as well as longtime auto industry consultants Daniel Bartz and Rich Chutorash, said one of the people familiar with the program.”
Justin Carlson’s LInkedIn backs up his being placed in charge. Miguel Hurtado previously worked at Valeo. Very little information about Rich Chutorash exists online. Daniel Bartz was previously at Peloton, the truck platooning company.
These all seem like sensible, industry-standard choices for FCA.
The rest of the article is similarly devoid of detail. Chrysler allegedly has prototypes on the road in Michigan, but didn’t respond to a request for comment. They’re allegedly using the same hardware as everyone else — radar, LIDAR and cameras — except Tesla, who eschew LIDAR. No surprises here.
“So many companies are working on developing self-driving technology that will allow automakers to buy the core pieces from other suppliers.”
I think that was an incomplete sentence. Omit the period, add “that” at the end, and it more logically flows into:
“The carmakers, including Chrysler, will be “the prime contractor who owns the integration [of self-driving technology with the car] but they may or may not own any of the pieces themselves,” said a person with knowledge of Fiat Chrysler’s plans.”
Now it gets interesting.
Automakers generally own the hardware that goes into their cars. The software? The Mobileye cameras that dominate the market are funny. Buying the camera doesn’t necessarily get you into the camera. Pulling raw data off their hardware isn’t easy, which is allegedly what led to their split with Tesla. Sources tell me Tesla wanted all the data (and a higher resolution camera) and didn’t want to share it or their Autopilot-derived logic with Mobileye, which was the latter’s condition for opening up their hardware.
And that battle was over a camera Tesla owned.
If FCA or anyone else is comfortable installing self-driving hardware they don’t own in their cars, then logically, they won’t own the data coming in through that hardware either. They may be able to use it on a limited basis, but if hardware is commoditized, software is everything. Actually, data is everything, because without data, software is only as good as the last line of code, and code gets old fast.
Garbage in, garbage out.
The core of Tesla’s strength is the massive amount of data gathered by their 175,000 cars on the road, whether or not they’re on Autopilot. This is what everyone wants yet no one but Waymo is in a position to gather. If Waymo can get two or more automakers on board through creative licensing agreements, they can ramp up their data gathering through a growing pool of users, strengthening their platform for all licensees.
The Real Secret
The Information’s source then all but admits that FCA is willing to give up the baby because their eye is on a different ball: the customization of self-driving behavior by brand and model, with a Dodge Challenger having more aggressive modes than their Chrysler sedans.
If Chrysler’s self-driving strategy is to convince customers a Dodge Challenger will be fun to ride in while aggressively driven by an AI, FCA is doomed. Self-driving behavior — as defined by software — will become as commoditized as word processors and spreadsheets. This is not a path to sales or profit.
If people want to be pinned to their seats and feel real cornering, they’ll want to be in control, even if it’s to their detriment. I know a lot of Challenger owners. Not one wants to give up control unless they’re being cradled like a baby, and that’s only in traffic.
Someone at FCA has made a mistake. Never bet against human nature.
There are only two metrics for measuring the “performance” and desirability of self-driving cars: safety and comfort. If your products aren’t clustered with the best, you’re going out of business.
The rest is cosmetic.
The Bottom Line
I believe the FCA source is telling the truth about their self-driving car project, primarily because their reasoning is so misguided, it has to be true. FCA, never having bothered to seriously invest in self-driving tech, is in the process of giving in to Waymo’s demands for data, and hoping to double down on brand by customizing self-driving behavior to match human driving behavior. That logic deserves its own article, but you don’t need to be Nostradamus to know that reeks of old thinking. If fourteen car companies will become six, the eight that don’t own a piece of the new technologies or networks that will soon dominate the marketplace are toast.
So why has FCA been so secretive? Everyone knows they've been looking for an acquisition, so perhaps this story was deliberately leaked so that a potential suitor would think they had a plan. They do. It’s not a good one. Alas, for someone in their position, it’s probably the only one. FCA’s a lot more attractive with the smallest possible self-driving car program. They should keep it that way. It's too late to do anything else.
By the way, there is a monumental opportunity for FCA (and everyone else) in semi-autonomous technology short of self-driving. Why hasn’t anyone seen it? More on that #soon...
Alex Roy is an entrepreneur, Editor-at-Large for The Drive, author of The Driver, and set the 2007 Transcontinental “Cannonball Run” Record in 31 hours & 4 minutes. You may follow him on Facebook, Twitter and Instagram.