User:Chrisvls/draft2

Officials from the City and County of San Francisco have reported that autonomous vehicles frequently interfere with fire department operations. In addition, they report frequent incidents of stopped autonomous vehicles causing access or traffic issues. These officials have argued that performance improvements should be required before expanding use of autonomus vehciles.

Officials from the autonomous vehicle companies have responded by citing the hundreds of lives lost on U.S. roads daily. Since safer autonomous driving could, in theory, save those lives, they argue that delay poses a greater risk to life than requiring product improvements or slowing the expansion of the current experiments with autonomous vehicles..

This note explores these argument by estimating these two offsetting risks: the risks increased by AV interference with emergency services versus the risks of accident deaths decreased by autonomous driving safety.

Deaths potentially prevented by autonomous vehicle safety

In 2021, 27 people were killed in automobile accidents on non-highway streets of in the City and County of San Francisco. Waymo and Cruise AVs operate only on non-highway streets. Selecting 2021 matches most recent the road usage data available to give us a risk of death as a function of road usage. The number of deaths per year have varied widely, however, with 2022 seeing much higher, 39 deaths, and 2023 being on track so far to be much lower, 13 deaths through July 31, which would indicate about 22 deaths for the year. The averages of the last five and nine years (not including the optimistic 2023), is 29.6 and 29.1, respectively. This, selecting 2021 data seems a reasonable choice, close to an average year.

In 2021, total non-highway roads traffic in San Francisco totaled 2,880,450 vehicle miles traveled per day, or about 1.051 billion miles in the year. This estimate comes from the California Department of Transportation.

So in 2021, we saw 27 deaths occur after 1.051 billion vehicle miles traveled. This yields the fairly straightforward math of 25.7 deaths per 1 billion vehicle miles. Put another way, to prevent one death, assuming that AVs have a perfect fatality record, we would need the AVs to replace approximately 39 human million vehicle miles.

Thirty-nine million miles is a significant increase over current driverless AV operation, as reported to the California Public Utilities Commission. In May 2023, Waymo and Cruise reported passenger operations (which includes driving miles waiting for a passenger) of [x] and [y] vehicle miles traveled, respectively, in May 2023. These figures would annualize to between 0.8 and 1.5 million miles traveled per year. These totals are quite variable, however.

Frequency of autonomous-vehicle interference with emergency servicesFrequency of collisions with a responding emergency vehicle

On date, a Cruise autonomous vehicle collided with a fire engine that was en route to a call. The accident injured the AV's passenger. While we only have systematically collected data of fire depratment interference for a few months, we have collision reports for AVs going back much further. Since this appears to be the only known actual collision with an emergency vehicle, it appears that this is a significantly more rare event.

Given

Frequency of stopped autonomous vehicle incidents

Emergency scenarios

The above analysis calculates frequencies and risk on the assumption that emergency response events are normally distributed random events. But this assumption does not apply during large-scale emergencies, such as a major disaster or earthquake.

Two recent events of note appear in the data: a significant storm in April 2023 and the Outside Lands Music Festival in August 2023. During the April storm

There are some scenarios in which