As fun as it is to get a ride in a self-driving car (even if it does have a safety driver) there are only so many times that is exciting enough to write about. So, for a change, Aptiv, a leading automotive component supplier and innovator in self-driving technology, offered me a behind-the-scenes tour of their Las Vegas Technical Center instead of another test ride. Their LVTC is responsible for both the overall operation of the extensive fleet of autonomous vehicles it operates with Lyft in Las Vegas and working on the portions of the underlying technology.
To put this in context, for nearly two years Lyft customers have been able to opt-in to getting a self-driving car to ferry them around Las Vegas — in lieu of a more traditional vehicle. So it is a real, commercial service. It is limited to certain drop-offs and pick-up points (3,400 currently, with Aptiv announcing here at the show that it will begin testing the addition of the airport to that list), but is one of the largest and longest-running commercial deployments anywhere.
Aptiv Garage: If Taxi Companies Invested This Much, We Might Not Have Uber
With room for 130 vehicles, the air-conditioned, spotless Aptiv garage is a testament to the level of investment autonomous vehicle companies are making in operations. It includes a full-calibration lab space, where vehicles are placed on a turntable and have their 20 or more lidar, radar, and camera sensors calibrated every 6 to 8 weeks or after the vehicle suffers damage or hits a large pothole. That’s something to think about before you get too far ahead of yourself wishing for a Level 4 car in your garage. Stalls all have drops for data feeds to pull the massive amount of information generated by each car into the site’s 5-petabyte data center and out through a 10Gbit data link to the company’s other R&D centers as needed. A row of car chargers stands ready for when the new hybrid Pacificas enter the fleet.
Aptiv’s Operation Center: High-tech Fleet Management
Backing up the fleet of 30 or more self-driving cars on the road — 2 shifts a day, 7 days a week — is a state-of-the-art operations center that monitors and can help position the cars. When I visited, there were about 10 people and about 30 computer and TV screens. I wasn’t allowed to photograph there, but the wall-size TVs showed current statistics, a map of locations of the current fleet, and from time to time a real-time video feed from the dashcam or dashboard readout of a specific car. Behind the operations center are extensive employee facilities, including a training center where safety drivers spend half of their first 6-8 weeks with the company learning the ropes (the other half is spent learning and training on the road).
Generations of Aptiv’s Autonomous Driving Research Vehicles
On display at the Tech Center are several generations of Aptiv’s test platforms. Each one is sleeker than the previous one. The Audi-based cars were the first to drive coast-to-coast nearly autonomously (Aptiv says 99 percent). They gave way to the BMW platform that demo riders are familiar with. Impressively, both vehicles look more-or-less normal, while hiding 9-10 lidar, 9-10 radar, and multiple cameras. By using multiple lidar, they can all be tucked nicely away inside the outline of the car. No KFC bucket on the roof. There’s even a cool hidden radar on the BMW that reveals itself in situations when cross-traffic is particularly important.
I was able to see, but not photograph, Aptiv’s next-generation Pacifica-based platform. Like many other AV efforts, Aptiv chose the Pacifica for a couple of solid reasons. One, seemingly-trivial, is that the doors are all automated. No need to dispatch a tech to close the door after a forgetful passenger. And the Pacifica offers redundant control systems, which Aptiv considers a safety essential. The Pacificas will be tested this year, but Aptiv didn’t have a date for when they might be picking up passengers.
Aptiv’s Autonomous Joint Venture With Hyundai
There is no lack of enthusiasm within Aptiv for self-driving research, so it might be surprising that the company is spinning off its entire Level 4 – Level 5 AV program into a half-owned joint venture with auto giant Hyundai. I was able to spend some time with the CEO of the new company, Aptiv Mobility, Karl Iagnemma, and ask him about the joint venture and its future. For starters, he explained why the two companies teamed up. While Aptiv has been working in the area for over a decade, each new generation of test vehicles has to be essentially hand-crafted. To get to scale they needed an automotive partner that could create the necessary autos around the world. Hyundai’s 7.5-million-vehicle-per-year global presence fits that need well. So does its checkbook, which brings $2 billion ($1.6 billion of it in cash) to the joint venture.
In exchange, Hyundai gets a huge leap forward in its future technology stack. But Iagnemma stressed that the joint venture is an independent company, without any exclusive relationships with its parent companies, and will be looking for partners and customers across the industry. As far as timeline — always a thorny question for the self-driving industry — he sees fleet deployments continuing to scale. Aptiv is planning to start testing truly driverless vehicles on city streets later this year, followed by larger fleets of ride-hailed vehicles, but he doesn’t expect the component cost of the needed sensors and compute to be practical for consumer vehicles until around 2030. I also asked about the state of the company’s Lyft partnership. They told me that while Lyft was opening up its effort that Aptiv still hoped they would be a customer in the future.
Along with the timeline, one of the other tricky questions for the industry is the choice of sensor technology. I asked Iagnemma how he saw the sensor array evolving over time. Currently, he sees the need for multi-modal sensing (in this case lidar, radar, and cameras) to continue in the foreseeable future. Both for redundancy, and for complementary features in a variety of driving conditions (lidars can’t see red lights, for example, but cameras can’t see things that aren’t lit, etc.) One additional sensor type that the Aptiv fleet uses (besides the obvious GPS) is a pair of sensors that can listen to the traffic lights at 120 Las Vegas intersections. They serve as a fail-safe for the camera that is used for red-light detection.
So Are Level 4 Fleets Really the Future for Urban Ride-Sharing?
Given the billions invested by each of at least a dozen companies — Cruise investors alone have poured over $5 billion in current and committed cash along with promised services into its test fleet — the industry clearly foresees a time when fleets of Level 4 vehicles roam the streets of cities worldwide. I did my best to tease out some statistics from my Lyft tour guide to see how well fleets are doing in the real world. He said that vehicle efficiency was similar to regular, driver vehicles, and anecdotally, that trips using their self-driving cars weren’t noticeably slower than those with drivers (which is one knock on Cruise in San Francisco, where avoiding certain intersections and routes can add a lot of time to non-driver trips).
So far, Aptiv’s Lyft fleet has completed just over 95,000 trips in Las Vegas (the number was prominently displayed in the operations center). One of the things they’ve learned is that passengers are very curious about the technology, and that one of the major jobs of the second safety driver is answering passenger questions. They also said that passengers have been uniformly positive about the experience, although of course, it isn’t really a true driverless experience since they’re in the car with two drivers. My overall impression is that Aptiv Mobility is a well-thought-out, well-positioned effort to create practical fleet-based vehicles suitable for finite areas that can be mapped and monitored. It also positions both Aptiv itself and Hyundai for the eventual day when Level 4 autonomous vehicles become practical for consumers.
- The Self-Driving Industry Is Finally Becoming More Realistic
- How Self-Driving Cars Work, and When They’ll Get Real
- Autonomous taxis: Why you may never own a self-driving car