Some Context About The Cruise Robotaxi Suspension
The suspension and subsequent temporary cessation of Cruise Robotaxi service in San Francisco and other cities across the U.S. is on people’s minds. Some want to have known all along that autonomous cars are not safe and have been unleashed on the public far too early, others don’t understand the excitement, given 43,000 traffic deaths each year in the US alone. Germany, for example, had more than 2,700 traffic fatalities in 2022, and Austria will have more than 300.
The Sequence
So what exactly happened that led to the suspension? In mid-October, a Cruise was driving parallel to a Nissan driven by a human. In the process, the Nissan driver ran over a female pedestrian who had crossed the street on a red light. The woman was then thrown under the Cruise robot cab, which immediately braked and came to a stop. The Nissan driver committed a hit-and-run and has not yet been apprehended.
In other words, the Nissan driver was entirely to blame for the accident. But what happened next has been occupying the public ever since. And this can be divided into the following: action of the Cruise Robotaxis, the authorities’ confusion, and the previous incidents.
1. Action of the Cruise Robotaxi
After the accident, the Cruise had come to a stop. The woman had remained trapped under the vehicle, which the Cruise was unaware of due to the lack of sensors on the vehicle’s underbody. Therefore, after detecting the impact of the pedestrian, the vehicle attempted to reach a safe location and drove several meters (20 feet) to the side of the road to avoid blocking and endangering traffic. The woman was dragged in the process, and the Cruise came to a stop with one wheel on the woman’s leg. Only the fire department that was called in was then able to free the woman from her predicament.
What is interesting to know here is that the Cruise behaved exactly as the authorities had requested. In the past, there had been repeated reports of stalled robotaxis blocking traffic. As a result, authorities have requested that the companies require the vehicles to move to a safe location. An example of the vehicle trying to move to a safe and non-blocking position was captured in this video:
Incidentally, this incident also clearly shows why it is necessary for robotaxis to be tested in reality. This accident represents one of those rarely occurring traffic events that are difficult to predict and test. Of course, one wishes that there would never be any personal injury, but, as I noted once before in another blogpost, unfortunately you just can’t build the best airplane by developing and testing it exclusively in a wind tunnel.
2. Authorities’ Confusion
Cruise received approval (together with Waymo) on August 11, 2023 to operate driverless robot cabs in commercial service, initially in San Francisco. The granting authority is the California Public Utility Commission (CA PUC), which is responsible for cab licenses, among other things. Even before that, the California Department of Motor Vehicles (DMV) issued test permits. Currently, about 40 companies have such a permit in various restrictions. The city of San Francisco itself, as well as the local police or fire departments, only have party status, but no say in the matter. This was a point of criticism by the city in the permitting process.
At the federal level, there is the National Traffic And Highway Safety Administration (NHTSA), which issues federal regulations, and the National Traffic And Safety Board (NTSB), which is responsible for investigations and final reports, especially in the event of accidents.
As you can see, the competencies are distributed and in some cases they overlap. And this becomes important when we compare the statements of individual authorities. For the DMV accused Cruise of having withheld the recordings of the Cruise – which, after all, has two dozen cameras and a number of other sensors – with the subsequent action of the authority. Cruise, on the other hand, claims the opposite, that it provided all the video footage to the authorities.
Here, it is not entirely clear whether all agencies received the same video footage, whether there were specific requests for the video footage, and whether video footage was intentionally or unintentionally withheld.
The DMV, in any case, viewed it as intentional withholding and revoked Cruise’s license, and the CA PUC followed suit the next day. Waymo, the operator of the other fleet of robotaxis, is unaffected and may continue to operate its driverless robotaxis in San Francisco and Los Angeles.
3. Prior Incidents
Due to a series of other incidents in the months since the license was issued, Cruise in particular had come under fire. From stuck vehicles, to collisions with a fire truck, a collision with a truck, and a blocked emergency vehicle that was then allegedly too late to help the accident victim (who had been hit by a bus and subsequently died), every incident, minor or major, has come into the public eye.
At the same time, initial statistics, still to be taken with caution, show that robotaxis are 50 to 75 percent safer than human drivers. For example, both Swiss Re’s figures with Waymo and Cruise have used two university studies to show these facts. They should be taken with a grain of salt because the two fleets have only driven a few million miles driverless, while the comparative figures for human drivers amount to hundreds of billions of miles. Also, a number of data are missing because not all incidents involving human-driven cars are reported – it is estimated that 1/3 are never reported – while robotaxis are currently required to report every incident, no matter how small, at the request of authorities.
Similarly, previous reports of incidents involving robotaxis by emergency responders – who have been specifically encouraged by their supervisors to report any incident, no matter how minor – showed that three-quarters of them were insignificant. Incidents involving human drivers appear to be much more common, but are reported only when they reach a certain level of severity.
Open Questions
The authorities’ approach and Cruise raise some questions.
Why was the DMV able to revoke the license first, when it was actually the CA PUC that issued it?
Why did the CA PUC follow suit a day later?
Who really has all the sovereignty over licensing now?
What exactly are the steps involved?
Are the same criteria applied across agencies here or are there differences?
Why is the license for the entire fleet of robotaxis revoked, while for human drivers, all cars in the same category are not immediately shut down?
On the last point, it might also be worth adding that for a century there have been repeated deaths and injuries because drivers in reverse gear overlooked someone and ran them over. In the USA alone, for example, 200 people died each year as a result. Despite this, rearview cameras were not passed as law in the U.S. until 2008 and have been required to be installed in all new vehicles since 2014. But the technology has been around much longer, and yet despite 200 deaths a year, no cars or entire vehicle categories have been shut down because of it.
This raises further questions as a consequence of the suspension:
Why are different measures used here?
If robot cabs are already safer today (see the statistics mentioned), won’t a suspension hurt society as a whole because there will be more deaths and injuries because human drivers will be called upon again?
Conclusion
Clearly, I would like to say here that of course a solution must be found so that such an accident does not happen again and a robot cab drags a person along. Sensors under the car floor and new algorithms are one approach, another could be an interaction with the vehicle by voice, where an accident victim could draw attention to himself in this way.
Unfortunately, new, unexpected, unpredictable and surprising incidents will continue to occur in the future, the prevention of which will then need to be incorporated into the robotaxi fleet accordingly as an improvement. And thus make the robotaxi fleet successively better. Quite in contrast to the totality of human drivers, who never get better overall.
Also, authorities need to learn how to deal with robotaxis better, develop better criteria for safety assessment and requirements. Also, at what point an entire fleet needs to be shut down and when not. So the current approach looks a bit like an overreaction, meant to satisfy critics, but could cause harm to society as a whole.
It must also be mentioned in Cruise’s direction that perhaps a more cautious approach in case of doubt would be better. I myself have been on the road with Cruise Robotaxi driverless about 130 times and have about 20 trips with a Waymo as a comparison. Cruise does tend to have a slightly more aggressive driving style than Waymo, although I never felt unsafe.
However, there’s no stopping robotaxis and autonomous vehicles, and what we’re seeing here are still the bumpy beginnings where authorities, companies and the public have to get used to each other and learn to understand each other.
This article was also published in German.