Uber, Wayve and Stellantis join forces to progress robotaxi technology

This week’s technology agreement announced between Stellantis, Wayve and Uber is another big step forward in the race to provide – and profit from – driverless taxis.

For once the story seems less about progress in the nuts and bolts of how the cars weave their way around road hazards. It’s more about how the next stage will be financed.

Behind pretty much every technological advance we report on, there’s someone financing it, hoping to profit. But when three global giants from completely different industrial spheres sign a ‘Memorandum of Understanding’ to deploy Level 4 driverless robotaxis around the world, it seems a logical corporate union. Three different specialists are merging their expertise to mutual advantage. Or are they?

Looking behind the handshakes and smiles, there seems to be an imbalance in who stands to gain most.

Uber is successfully outsourcing the expensive, high-risk elements of the self-driving war – and has perfectly positioned itself to collect massive anticipated profits. No paychecks for drivers mean higher yield per journey. Once you’ve bought the autonomous cars, you sit back and collect the money. Obviously Stellantis and Wayve aren’t purely in the business of providing profits for Uber but let’s look deeper at what’s really happening.

Current Level 4 autonomous driving systems rely on highly detailed, pre-programmed, high-definition mapping of a defined geographic area. These are carefully compiled for each city, meter-by-meter, junction by junction, before a car can operate there. It means rival operation Waymo’s robotaxis are trapped within rigid geofences.

This new partnership is taking a different technological route. British autonomous-driving-tech company Wayve is devising a "mapless" self-learning AI system allowing a vehicle to navigate in areas it has never previously visited.

Wayve's "Embodied AI" system processes real-time data through a vehicle’s cameras and radar sensors and uses advanced machine learning to interpret and adapt to these brand-new, complex environments without any pre-existing mapping. It means Wayve’s technology should eventually be able to operate across different regions exponentially faster and cheaper than Waymo.

In the new trio, Wayve is clearly supplying the AI "brains." And it seems that Stellantis is taking on the monumental financial risk of providing the "brawn." The giant automotive manufacturer’s job will be designing, engineering and mass-producing the autonomous vehicles. These, it says, will be built on its specialized "L4-Ready Platforms."

This means Stellantis has abandoned the idea of retrofitting cars with bolt-on sensors, computers and cameras. Instead, it’s integrating bespoke sensor systems into a chassis engineered specifically for driverless operations and tough enough for a life as a 24/7 Uber.

So, what does Uber bring to the table? The taxi-app company has already abandoned its own expensive self-driving vehicle program. With this new agreement, Uber will be able to simply deploy Stellantis cars with Wayve’s AI into its existing taxi network. Passengers will book an autonomous trip via the standard Uber app. Uber does not have to spend billions developing AI or vehicles. Stellantis and Wayve tackle all the R&D and hardware for it. Uber will apparently rewrite its app slightly and start collecting the taxi fares. It looks like a very good deal for Uber.

Here’s a little background to the tech being discussed here. Level 4 autonomous driving means the car does all the driving without any human help. You can read or sleep as the vehicle motors along. However, it only works in specific, pre-mapped areas or good weather. If a problem happens, the car safely pulls over by itself instead of asking you to take over.

Level 5 autonomous driving is the higher goal: complete, unrestricted automation. And that’s what the Wayve-based system is hoping to offer Uber. It means the car handles all driving in any location, on any road and in any weather, just like a human driver.

Level 5 can navigate unpredictable dirt roads, unmarked rural lanes or chaotic city centers anywhere in the world. It can handle extreme blizzards, torrential downpours or heavy dust storms – all potential problems for Level 4.

Because a human never needs to take over, Level 5 vehicles do not require a steering wheel, accelerator pedal or brake pedal. The entire interior can be redesigned into a mobile living room or office.

The robotaxis already carrying passengers in specific US cities are all Level 4. Level 5 technology does not exist on the road yet. Achieving it will require artificial intelligence that can think, adapt and reason through completely unpredictable real-world scenarios just like a human brain.

The crucial point comes with an unexpected, highly unusual scenario. While AI can be trained on billions of miles of normal driving, that doesn’t help when the unexpected arrives, like a police officer using improvised odd hand gestures to direct traffic, a sudden emergency construction zone with hand-painted detour signs or a horse-drawn carriage sharing a rural road in a sudden downpour.

Humans deal with these situations using common sense but current AI can freeze when encountering a situation not present in its training data. A perfect example was recently revealed at an industry conference: an autonomous car performed a sudden emergency stop in the middle of a busy road after mistaking a life-size movie poster on the side of a bus for real people standing in the road. Creating AI that can avoid mistakes like that is Wayve’s challenge.

Source: Stellantis