Number-two ride-sharer Lyft and automotive OEM Aptiv (formerly Delphi) have announced that their 20-car fleet of self-driving taxis have now served 5,000 trips in Las Vegas since January. It comes ahead of Waymo’s plan to start charging fees for its service in Arizona. When viewed in combination, these trials could create the consumer and political trust needed to bring fully autonomous services to market.
Without such trust or buy-in, the likes of Aptiv and Waymo will struggle to find customers for their technologies. If, for example, a deadly school-bus crash was blamed on a self-driving car, a politician might seize that as a chance to push an agenda, playing on the tragedy to outlaw such cars on public roads. That would be disastrous for self-driving stakeholders, and they know it.
Consequently, they will be very cautious in these early stages, especially after Uber showed the world exactly how not to carry out such testing. Waymo has an exemplary record, and having the largest fleet and highest amount of test miles, seems to have set the benchmark.
Inner-city testing such as these taxi services also lowers the risk of being involved in a deadly crash. Lower average and top speeds inherently means less chance of injuries and fatalities, if a car is involved in a crash. There’s also the benefit of exposing the car to more of the most unpredictable things on our roads – pedestrians.
So Aptiv and Lyft’s project in Las Vegas, Cruise’s tests in San Francisco, Voyage’s pilot in Florida, and Waymo’s plans in Arizona, should collectively demonstrate the safety of these self-driving systems – while also training up the cars to be better drivers.
However, the biggest step forward will be when these taxi services ditch the test-drivers entirely. Waymo has taken some initial steps in that direction, providing some completely driverless trips in Arizona since October 2017. However, it will make history when it starts charging for these trips in Phoenix, Arizona.
The Verge spent a day in the Waymo depot, profiling how the company is getting to grips with the fleet-management aspect of the service. Pre-shift checks are carried out, calibrating navigational sensors and performing the routine maintenance and inspections on the mechanical aspects of the cars, before they head out for a shift.
These shifts are described as ‘fairly mind-numbing’ for the backup drivers, and it appears that the shift lasts around six hours. The fleet is servicing the Early Rider program, an opt-in list of around 400 people in the locale of the depot, who can request trips – which recently kicked into 24/7 operation. Waymo’s total fleet is now over 600 vehicles, but it’s not clear how many are in Arizona.
In terms of expansion, Waymo has reportedly signed deals to procure up to 62,000 Chrysler Pacificas, and 20,000 Jaguar I-Paces. New cities will be added, with strong candidates being Atlanta, Detroit, Mountain View, and San Francisco. It still plans to license its IP to automakers, as well as launch a standalone ride-sharing service.
A central remote team is monitoring the fleet’s aggregate performance, and issuing updates to the fleet. Similarly, that team can experiment with different parameters, to see what happens to the performance of the fleet. With plans to start serving anyone, not just the Early Riders, the ability to put the cars in the places where they will be needed is a top priority.
This team also responds to the car’s requests for additional help, when they encounter an unfamiliar problem on the road. The solution to those problems can then be pushed to all the other cars in the fleet, which is a practical, if intermediate, step towards full autonomy. Notably, this team can’t directly control the cars, and are essentially acting as advisers. A local team is on-hand to respond to customer requests, and can use cameras to check cars for things such as missing items or messes.
The Lyft-Aptiv fleet if part of a 75-car fleet that Aptiv keeps in Vegas. Aptiv is planning to increase the taxi fleet to 30 cars in the next few weeks, as well as expand its destination offerings in downtown Las Vegas.
In terms of structure, Aptiv is responsible for the car, and Lyft manages the client-side elements. Essentially, a Lyft user will be asked if they’d like to participate in a test drive when they open their application to book a trip. If a user accepts and opts-in, then the car will be sent to pick them up. The revenue for the trip is then split between Lyft and Aptiv – although neither company has disclosed how that is divided.
For the duration of the trip, the backup driver is on-hand, incase they need to intervene. Unsurprisingly, there’s been no disclosure about how often they have needed to step in, but nor have we seen a video of this happening – which we’re sure would bubble to the surface of the online press, if it existed. Thanks to Lyft’s platform, the two companies are able to track customer feedback, which so far sits at an average of 4.96 out of 5 stars.
“One of the things that’s really different about this for us is that we think of this as a scaled program,” said Jody Kelman, product lead for Lyft’s self-driving platform, to The Verge. “In starting to charge our passengers for rides, that’s really for us a notable transition into something we want to continuously operate on the Lyft network as a real part of our transportation operation.”