Waymo: Showdown at SFO

As part of its financial filings in 2018 when it was preparing to go public in 2019 Uber noted that 15% of its gross bookings were for trips that either began or ended at an airport.? More specifically, Uber reported that 24% of those airport-centric ride-sharing gross bookings came from five metropolitan areas – Los Angeles, New York City, the San Francisco Bay Area, London, and Sao Paulo, Brazil. That translates to hundreds of millions if not billions of dollars today.

It is therefore no surprise that Waymo is focused like a laser on obtaining approval from the California Public Utilities Commission and the Department of Motor Vehicles for commercial operation of its robotaxi service to be extended to include SFO. ?Waymo is currently operating 100 robotaxis within a portion of the City of San Francisco.? The company says it is offering “tens of thousands of fared rides per week to members of the public in San Francisco,” according to its most recent filings with the state.

Simultaneous with Waymo’s efforts to obtain this approval the State legislature is considering State Bill 915 which would give approval rights to local authorities.? Called the Autonomous Vehicle Service Deployment and Data Transparency Act the bill is intended (according to its authors) to:

• Empower Local Communities: Instead of the State of California determining how self-driving vehicles operate in a given area, local governments will be empowered to grant such permits, as well as enforce traffic laws on autonomous vehicle companies by enacting new local ordinances, as needed, in real-time.
• Maintain Safety for Pedestrians, Passengeres, and Other Vehicles: Local communities will ensure that self-driving car services are safe and user-friendly for everyone, including passengers with disabilities. Tech companies will need to demonstrate that each of their autonomous vehicles are regularly inspected to ensure safety. Pedestrian-populated areas like schools and airports can be made safe by local ordinance.
• Establish a Community-Centered Approach: SB 915 allows each local government to set up its own rules for self-driving vehicles based on the community's unique needs. Local governments will be able to govern fares on robotaxis and the number of autonomous vehicles on the road. This will manage traffic and reduce the potential for injuries.

At stake for Waymo is access to the lucrative airport traffic currently dominated by Uber and Lyft along with traditional taxis.? These are longer trips, generating greater gross revenue, capable of catapulting Waymo closer to profitability and viability.

These trips also will require Waymo’s vehicles, uplifted electric Jaguar I-Paces, to operate at higher speeds on Route 101 across multiple local jurisdictions.? Waymo’s filing proposes an operating area from San Francisco southward to include Sunnyvale, with the western boundary running along Highway 280 and the eastern boundary mostly along the Bay.

According to reporting in The Detroit News, “Waymo said in its proposal that the DMV has approved its robotaxis for use day and night on freeways and highways with speed limits up to 65 miles per hour, along with city streets, rural roads, other roadways and parking lots. The vehicles … would be eligible to operate in rain or fog, but not snow, according to Waymo.”

In essence, Waymo is seeking to expand what is known as its ODD – operational design domain – i.e. the areas and conditions for which it is suitable to operate and for which its hardware and software systems were designed.? Waymo is clearly identifying speed, geographic range, roadway types (including freeways and parking lots), and weather conditions with which its vehicles are able to cope.

It seems clear that neither the California DMV nor CPUC have the testing or regulatory protocols in place upon which to base an approval of Waymo’s request.? Additionally, local authorities throughout the defined area including the Airport have gone on record to oppose state-level approval of Waymo’s robotaxi operations in their jurisdictions.

These developments serve to highlight Waymo’s collaboration with insurer Swiss Re in developing and publishing a report last Fall – “Comparative Safety Performance of Autonomous - and Human Drivers: A Real-World Case Study of the Waymo One Service.” ?The study concluded “that the Waymo One autonomous service is significantly safer towards other road users than human drivers are, as measured via collision causation.”

The study broke new ground in seeking to make “apples to apples” comparisons, in the words of one of the report’s authors, between the 3.8M autonomously driven miles of Waymo One vehicles vs. many more millions of miles driven by humans.? In making its comparison, Swiss Re developed “human driver baselines based on Swiss Re’s property damage liability and bodily injury liability claims data from 2016 to 2021, from over 600,000 claims and over 125B miles of exposure.”

Swiss Re specifically focused on insured drivers drawn from the area codes where Waymo is currently operating.? In a presentation at an AutoSens event last fall, a senior Waymo official detailed the various challenges of achieving a like-to-like comparison of Waymo and human drivers.

Those challenges included using Swiss Re data, since reported crash data of the sort available from the U.S. Department of Transportation (Fatality Analysis Reporting System (FARS) or the National Highway Traffic Safety Administration Standing General Order for AVs) or California DMV (California Code of Regulations 2014) were both inconsistent and incomplete.? Hundreds of thousands of vehicle-related incidents generate insurance claims without being reported to state or Federal authorities.

According to Waymo, the vectors of data discrimination break down as follows:

1. 60% of property damage collisions and 32% of personal injury collisions are not reported to police; there is an unknown number of <$1,000 property damage and no damage events that go unreported
2. Broadly speaking, the lower the severity of the incident of vehicle contact, the less likely it is to be reported
3. Crash data per million miles traveled needs to be corrected for the fact that the average number of vehicles involved is 1.8 – not 1 – which increases the number of vehicle crashes per million miles to 3.5 from 2.0.
4. ODDs must be consistent including taking into account time of day, weather conditions, and vehicle type

Waymo and Swiss Re also note that the claims data used by Swiss Re is more detailed, consistent, and complete compared to data reported to local or Federal authorities.? The report concluded that:

“In over 3.8M miles driven without a human being behind the steering wheel in rider-only mode, the Waymo Driver incurred zero bodily injury claims in comparison with the human driver baseline of 1.11 claims per million miles.? The Waymo Driver also significantly reduced property damage claims to 0.78/million miles in comparison with the human driver baseline of 3.26/million miles.”

Assuming this assessment is accurate even taking into account the various adjustments necessary to line up human vs. machine driven performance, challenges remain.? Robotaxis can be counted on to obey traffic laws and are not subject to the vagaries of human driving which might be vulnerable to emotion, impairment, drowsiness, fatigue, or distraction.

But robotaxis have demonstrated a vulnerability to weather and interactions with first responders and their vehicles.? The ability of Waymo vehicles to detect and avoid pedestrians and other road users has earned the support of cyclists – even though a Waymo vehicle hit a cyclist in the past week.

The collision with the cyclist, which occurred in San Francisco and resulted in non-life-threatening injuries, only served to highlight a further reality from the Swiss Re study.? Robotaxis are capable of detecting all incidents whether they cause little or no damage or require no reporting.

The Waymo AutoSens presentation noted: “In 1M miles (Victor, et.al ., 2023), Waymo included reporting of all 20 contact events, even with no property damage:

• 9 contacts with no property damage
• 9 contacts with scratches and/or dents
• 2 CISS-reportable (police report & tow away) collions

The exemplary safety record portrayed in the Waymo/Swiss Re report sadly is not able to anticipate how an expansion of Waymo’s ODD might impact the results of a future study. ?As ever, car insurers base their underwriting on historical data including claims and driving infractions.

Waymo is seeking authorization to expand its ODD to highways and airport access.? Highways not only present the challenge of higher speeds but also the need to change lanes and merge, the possibility of being hit from behind, to say nothing of the jockeying for position at the airport terminal curb.? Judging from various Youtube videos of Waymo vehicles in action, it is likely they will be up to the merging challenge.

Following the disruptions to San Francisco first responders caused by both Waymo and the now-dormant Cruise, local leaders throughout the geographic area being targeted by Waymo are expressing strong opposition.? The challenge for Waymo now is one of public relations outreach to first responders combined with some education, humble pie, and some suspension of disbelief on the part of those regulators and local leaders.

The Waymo-Swiss Re study is enlightening, but not conclusive given its backward glance in the context of the current ODD expansion.? At stake are lucrative airport runs that might well make the difference between profitable and loss-making robotaxi operations. ?If Waymo gains approval for SFO runs it will open the door for other robotaxi operators and serve to put real pressure on ride hail operators…and not just in San Francisco.