When Mike O’Shaughnessy, deputy director for operations at Phoenix Sky Harbor, learned of a plan to introduce robotaxis at the airport, he was anything but excited. “When I first heard about it, I was like, ‘I don’t want them on my roadways. It’s going to cause problems,’” he says. Hundreds of thousands of autonomous vehicle (AV) pickups and drop-offs later, O’Shaughnessy is an advocate.
Sky Harbor is one of the busiest airports in the country; last year, 53 million passengers passed through it. Drivers tend to be on edge, O’Shaughnessy says — distracted by cellphones, looking for their party, jockeying for the perfect space. “An autonomous vehicle doesn’t do that,” he says. “It’s focused on its mission.” The cars follow the rules of the road in and around the terminal. They obey speed limits. They come to a complete stop at stop signs. They don’t roll through yellow lights.
Autonomous vehicles were introduced at Sky Harbor in stages, starting in 2023. At first, they were driven by humans as the cars gathered data from radar, cameras and lidar (pulsed laser sensors) to learn about the driving environment. Next they were guided by computer systems, a human onboard in case of trouble. Then they were fully automated. Over a period of years, they were brought from airport outskirts into the main stream of traffic. Today they have 24/7 access, making more than a thousand pickups and drop-offs each day.
O’Shaughnessy says there have only been a couple of “very minor incidents.” Causes have been quickly identified and corrected. “I trust them more than I trust myself driving — that’s for sure,” he says.
Some expected this transition to be much further along by this time. In a book published in 1940 titled Magic Motorways, designer Norman Bel Geddes expressed optimism that highways could be filled with self-driving cars by 1960. “Human nature itself, unaided, does not make for efficient driving,” he wrote. “Human beings, even when at the wheel, are prone to talk, wave to their friends, make love, day-dream, listen to the radio, stare at striking billboards, light cigarettes, take chances.” Driving safety could be best achieved by eliminating the “human factor.”
The January 1958 edition of RCA’s Electronic Age described an “electronic drive” mode. The highway of the future would be embedded with wires sending signals to transistorized receivers in cars to guide a car accurately even under conditions of “zero visibility.” Drivers could turn their attention to conversation, work or reading. More recently, a 2017 report from the think tank RethinkX predicted that by 2030, 60 percent of the cars on the road would be self-driving “transportation as a service” vehicles.
The transition to an era of autonomous driving is just getting started, says Henry Liu, director of the University of Michigan Transportation Research Institute. It encompasses infrastructure, support, regulation, legal liability, labor and workforce issues. Researchers now believe the technology will reach full capacity in about the next 10 years. “Because the impact is so big, it will take time — it’s definitely not in the immediate future,” Liu says. “But it will happen, and I’m quite optimistic.”
Mcity
At present, autonomous driving technology company Waymo has by far the greatest share of the robotaxi market. It operates 2,000 vehicles in five jurisdictions — Atlanta, Austin, Los Angeles, Phoenix and the San Francisco Bay Area — and plans to bring them to nearly 20 more in the near future. To date, deployment has been concentrated in warmer climates. Snow can obscure roadway boundaries and lane lines; “black ice” and ice under snow can be invisible. Waymo began to test vehicles in Detroit in November, the first test of its systems in a snowy climate.
Warm Phoenix was the first city to implement Waymo robotaxi service, in 2020. Gov. Doug Ducey was instrumental in this, says Steve Zylstra, president and CEO of the Arizona Technology Council. Ducey invited Waymo to work with the Legislature and trade associations to create an industry-friendly regulatory regime. “Arizona is sort of the antithesis of California,” Zylstra says. “We have a very, very light regulatory regime and very low taxes.”
Zylstra sees multiple benefits from expanding robotaxi fleets, including reducing congestion and emissions. He’s been impressed by Waymo’s safety record. “These machines are so much better than humans,” he says. “They don’t make mistakes, and when they do it’s usually some kind of programming error that can be fixed through software.” But even after years of deployment, Phoenix is still ironing out kinks. A Waymo vehicle with a passenger in it recently went down a railroad track when a train was coming. (It got off the track without harm to the passenger and the software problem that led to the incident was corrected.)
Liu advocates third-party testing of autonomous vehicles. At present, local permitting doesn’t require independent certification of their “driving competency.” There are no federal requirements either, but a policy of “self-certification” by manufacturers. Liu directs Mcity, a public-private transportation research partnership at the University of Michigan. Eleven years ago it established the first “proving ground” for autonomous vehicles. The 32-acre test site includes physical infrastructure that reflects real-world driving challenges as well as “digital infrastructure,” data sent to vehicle control systems. This can test response to a wider range of driving events, including those involving pedestrians or other vehicles.
Waymo recently published a summary of the safety records of vehicles, which had driven 127 million miles in San Francisco, Los Angeles, Phoenix and Austin without a human driver. Compared to vehicles driven by humans, they had 90 percent fewer crashes resulting in serious injury and 92 percent fewer pedestrian-involved crashes.
More than half of all Waymo vehicles are in California, a thousand in the Bay Area alone. In response to the regulatory and public safety concerns raised by this influx, the San Francisco County Transportation Authority (SFCTA) developed a conceptual framework for autonomous vehicle permitting. It’s the first attempt to refine this process that Liu has seen from a local government.
The California Department of Motor Vehicles issues permits for autonomous vehicles in three stages: autonomous vehicles controlled by a driver; those that have a driver aboard for emergencies; and fully autonomous vehicles. But the state hasn’t actually adopted comprehensive permit requirements for each of these stages. SFCTA built out a framework for performance and safety metrics that it hopes the state will eventually adopt.
The framework “reflects our values and initial thoughts on what bench-marks, standards and metrics would need to be demonstrated to advance from one [permitting] stage to the next,” says Tilly Chang, SFCTA’s executive director. SFCTA suggests weather, speed and road restrictions as autonomous vehicles aim for a full permit, and requires the vehicles to meet safety targets. Cities around the country have come to SFCTA for advice.
Despite this show of commitment, public transit, biking and walking are still her “transit first” county’s biggest priorities, Chang says. She sees a role for AVs if they can work in harmony with others driving on congested Bay Area streets. “Technology is not for the sake of technology,” Chang says. “It’s always in service of your larger goals.”