Mike Ramsey, an analyst at Gartner who tracks autonomous-vehicle development, said it would not surprise him if Waymo achieved its ambitious timetable. But he added, “I anticipate a rollout that will be less than perfect and at times will be a bit slow and clunky.”
Waymo’s aggressive plans are all the more striking given the setback to another contender in the driverless push, the ride-hailing service Uber. One of Uber’s self-driving vehicles struck and killed a woman last week in Tempe, Ariz., in what is believed to be the first pedestrian death involving autonomous driving. As a result, its road testing has been halted in Arizona, California, Pittsburgh and Toronto.
“Self-driving vehicles will take some time to overcome the critical performance and safety challenges,” Bill Fay, Toyota’s senior vice president for automotive operations in North America, said Tuesday at an auto-industry conference hosted by J. D. Power and Associates. Toyota, like G.M., is setting a more deliberate pace in its driverless program.
And Waymo, as it moves into commercial operations, is likely to encounter questions that have not been addressed, such as how driverless cars will be insured, whether the federal government will regulate their use and what role the local authorities will play.
“The gap between testing fleets of vehicles and implementing them in real commercial fleets is going to be significant,” said Mr. Ramsey, the Gartner analyst.
Until now, Waymo has been using a fleet of more than 600 vehicles, most of them Chrysler Pacifica minivans, for its road tests. It agreed to use the new Jaguar I-Pace, a battery-powered hatchback unveiled this month, for the service announced Tuesday. After testing, the company expects to start its ride service by the end of the year.
Mr. Krafcik, the Waymo chief, said the company intended to move forward rapidly with a driverless ride service for the public because it was confident its vehicles could operate safely in virtually any driving situation where they would be put into use. In addition to the five million miles its cars have driven in tests, Waymo has run computer simulations equivalent to five billion miles of testing. In these computer models, Waymo uses software to make virtual test drives even more complicated by adding more pedestrians and making distances between cars shorter than in typical road situations, he said.
At a closed site in California, Waymo also tests the performance of its cars in traffic situations based on actual crashes. “We think we will be the safety benchmark for the world,” said Mr. Krafcik, who worked in operations at Ford and then was Hyundai’s United States chief before his arrival at Waymo.
He noted that Waymo cars were equipped with highly sensitive cameras, radar and three different types of lidar — a kind of radar based on lasers — to identify objects, pedestrians and other cars.
Even with the vast computing power of Google, which operates giant data centers around the world, Waymo won’t be able to train its computerized “driver” to be prepared for everything. “There are going to be things we can’t predict,” Adam Jonas, an automotive analyst at Morgan Stanley, said at the J. D. Power conference.
The Uber crash, at about 10 p.m. on March 18, involved a Volvo XC90 sport-utility vehicle outfitted with radar, cameras, and other sensors and computer gear to enable it to navigate without input from a driver. Elaine Herzberg, 49, was walking her bicycle across the road at a spot without a crosswalk when she was struck. Videos taken by cameras mounted in the Uber vehicle show the safety driver was looking down just before the impact.
The Uber crash prompted Toyota to suspend testing its self-driving cars on public roads as a precaution, although Ford Motor and General Motors said they were proceeding with their efforts.
Repeating comments he had made over the weekend, Mr. Krafcik said he believed that a Waymo vehicle would have been able to detect the pedestrian and avoid a crash.
He also clarified the business Waymo is pursuing, explaining that the company did not intend to make cars itself. It wants to develop self-driving technology that it can put into cars made by partners, and intends to operate them in a ride service.
Waymo also expects to put its technology into trucks and delivery vehicles, as well as buses or other vehicles that ferry riders to mass transportation systems. And it hopes to license the technology to carmakers, auto technology suppliers or other companies.
Mr. Krafcik said Waymo expected these efforts to become a highly profitable business, but declined to say how long that could take. Alphabet, the parent of Waymo and Google, “tends to take a very long-term approach,” he said.