2024-01-09
Waymo testing driverless robotaxis on Phoenix’s highways
Waymo just pressed pedal to metal in the race for full autonomous driving, launching the first-ever public robotaxi service capable of highway speeds without human safety operators. Initially available for employee use in Phoenix, Arizona, accessing freeways unlocks a crucial pillar to make self-driving cars ubiquitous.
"Highways represent the final frontier for operational domains," explained Professor Bart Selman, leading AI researcher at Cornell University. "Mastering complex maneuvers like merges, exits, and lane changes at 65 mph pushes the boundaries of perception and planning."
While Waymo logged over 20 million miles across various self-driving programs since 2009, this specific segment remained confined to test tracks. Safety drivers seated and ready overridden during trials involving unpredictable dimensions like extreme weather, complex intersections, and unusual obstacles.
Highway Hurdle Cleared - What's Next?
Now, Waymo has sufficiently advanced sensing, prediction, motion planning, and control to clear this highway hurdle without direct human supervision. How? The crux lies in ensuring smooth reliable handoffs between distinct operational design domains (ODDs).
"Highways themselves aren't profoundly different from surface streets," said Nathan Spielberg, Waymo Lead Machine Learning Engineer. "It's gracefully transitioning between them that creates complexity."
Imagine approaching convoluted cloverleaf interchange ramps with cars jostling for position. Sharp curves loom as the number of feasible paths shrink. Lanes splinter and merge at odd angles amidst forests of signage.
Here, Waymo leans on ultra-detailed HD maps combined with lidars granting 500 meter vision range - a major upgrade from early 25 meter systems. This enables adjusting speed, positioning, and signaling judiciously to slide into gaps based on downstream traffic flow and signs.
Waymo Also Capitalizing on Trucking Experience
And Waymo holds special advantage courtesy of its AV trucking division - recently reoriented towards robotaxis after learnings proved transferrable. Class 8 tractors tackle equally high velocities and hazardous maneuvers on freeways.
"Frankly, a 16 ton semi navigating rush hour Los Angeles traffic poses harder challenges than a passenger vehicle along relatively straight Arizona highways," Spielberg explained, noting how team talent transitions bore fruit.
The Next Stop is Scale
Starting in Phoenix, where Waymo One already operates the world's only live fully driverless taxi fleet over 280 km2, the addition of highways provides a blueprint expanding across sprawling Sun Belt metroplexes pending regulatory approval.
Interstate 10 bisects Phoenix, connecting Los Angeles and Las Vegas in the west to San Antonio and Houston in the southeast - both ripe robotaxi markets in Waymo's crosshairs.
Spielberg confirmed Waymo continues logging extensive additional ODD testing data in high-density San Francisco and Los Angeles to supplement existing operational domains.
Across all locales, highway access also makes ride-hailing routes far more efficient. Imagine returning from Phoenix's sprawling Sky Harbor Airport. Previously, a Waymo would be forced to ply congested surface arterials. Now marrying freeways slashes transit times through fast, relaxed cruising.
The Power of Software
While the lustrous self-driving Lexus RX and electric Jaguar I-PACEs personify cutting-edge mobility advancements, arguably Waymo's greatest differentiation stems from code.
Vault, the proprietary software stack, assimilates decades of structured deep learning across vision, behavioral modeling, prediction, planning, and simulation running smoothly 24/7/365 after fifteen years incubation.
Armed with ample processing muscle and burning terabytes of data daily, it makes billions of driving decisions reliable as human counterparts or better. This is the special sauce making full driverless highway speeds possible.
Not Resting on Laurels
"Reaching new operational milestones keeps energy levels high internally even after almost ten years perfecting our offering," said Spielberg. "Pushing limits and crossing into new territory feels like those early days at Google X."
The Cornell professor Selman agrees avoiding complacency remains important despite leads over rivals like Cruise and Motional.
"Expect the competition to keep heating up with well-funded efforts from the likes of Apple, Amazon, Baidu, Toyota, GM and Ford," he said. "Waymo's software and operational excellence does afford them distance, but constant innovation is essential."
Luckily with parent Alphabet's financial backing and Prius fleet scale benefits, Spielberg confirmed that's exactly the roadmap. Vault 4.0 architecture revisions are already underway alongside next-generation in-house lidars and radars.
Staying a few steps ahead ensures that whether in suburbs or skyscraper cities, Waymo customers can ride safely and smoothly on roads and highways at state-of-the-art speeds today and tomorrow. Full autonomy beckons.
Share with friends:
Write and read comments can only authorized users
Last news