Weekly Brief: While Uber Calls in the Lawyers, Wayve Trusts to AI on London’s Streets

The city of London was in the spotlight again last week when its transit authority revoked Uber’s operating license for jeopardizing customer safety.

This is the second time in three years that Uber has been kicked out of London for this same issue. Back in 2017 the first ban was overturned upon appeal. Given that London is Uber’s largest market in Europe, it’s no surprise that the ride-sharing giant has again appealed Transport for London’s decision. Will the ban stand this time? Probably not considering Uber’s access to expensive legal representation but our Paul Myles can get you caught up on all the details.

Meanwhile across town driverless car start-up Wayve deployed eight self-driving electric Jaguar I-Paces onto London streets. The company says it plans to conduct a trial of its autonomous tech throughout the UK but will concentrate its driving around Zones 1, 2 and 3 in central London. A bevy of venture capital firms from both sides of the Atlantic participated in a Series A funding that totaled $20M.

None of this is all that interesting on its own because every month another self-driving car start-up successfully completes another round of funding. Every month another self-driving car trial launches in London, New York, San Francisco, Beijing, or out in the desert in the company of tumbleweed and red sage. Another week, another trial, another twist of the hamster wheel that has become the driverless car industry.

However, it turns out that under the guise of a totally ordinary announcement that Wayve is up to something interesting, unusual and maybe even truly innovative here. For starters, those eight Jaguar I-Paces won’t have any LiDAR on board. Tesla can say the same thing about its Autopilot, which it relies on a mix of cameras, radar and ultrasonic sensors to accomplish autonomy. Yet Wayve’s driverless cars won’t even have any radar or ultrasonic sensors on board either. In fact, the only thing Wayve’s cars do have onboard to help the vehicle read its surroundings is a standard sat nav and a camera.

With that and, that alone, the company says it’s comfortable deploying its vehicles onto streets that they have never seen before, including environments abuzz with all the chaos of a crowded London street. That’s something that industry leader Waymo would never do.

The innovation behind Wayve’s trimmed down set up is artificial intelligence and machine learning. Wayve says its technology can learn to drive just the way a human learns to drive, with a mix of imitation learning, such as from watching our parents, and then reinforcement learning after we get a driver’s permit and practice over and over again. Usually all it takes is about 50 hours to become a competent driver who cannot just follow the rules but anticipate other drivers’ actions. Why should it take a driverless car company like Waymo 10 years, hundreds of millions of dollars and billions upon billions of driving data to create a vehicle that drives less confidently than a teenager after less than a week behind the wheel?

Wayve says that in place of hand-coded rules, it set out to build intuition into its self-driving technology. The reason that’s important is because no matter how many miles a self-driving car has driven, no matter how much data engineers have exposed it to and no matter how comprehensive the HD mapping, there will always be situations that don’t align with previous scenarios or data. An object might be the same, say a parked maintenance vehicle, but the lighting behind may be different or the weather may have changed. Unless self-driving tech can become intuitive, Wayve says, it will always be surprised by so called “edge cases” which can lead to either a slow and herky-jerky driving experience or worse, accidents.

The trial that Wayve launched in London last week is, therefor, one of the world’s first attempts to see if a driverless car can learn the art of driving rather than master the details of the road. If it’s successful, Wayve believes that it will be able to scale more safely and quickly than its competitors because it can enter any market without exhaustive testing beforehand.


Leave a comment

Your email address will not be published. Required fields are marked *