A joint study by The University of Michigan Transportation Research Institute and GM robotaxi subsidiary Cruise indicates self-driving taxis are involved in considerably fewer accidents than cabs with a human driver behind the wheel.
The study created a safety benchmark for human-driven taxis based on data from 5.6 million miles of ridehail driving information in order to compare human versus AV safety performance.
Overall, the study found human-driven taxis were involved in 64.9 collisions per million miles driven, while Cruise AVs were in 23 accidents per million miles. The human-driven taxis were in 32.5 collisions per million miles in which they were the primary cause of the accident versus 2 collisions for Cruise robotaxis, while 11.8 collisions per million miles carried a significant risk of injury for human vehicles versus 3 per million for Cruise.
The findings showed drastically lower rates of accidents and potential injuries for driverless vehicles compared to those with drivers. Data for both human-driven and self-driving taxis came from San Francisco and excluded accidents that occurred above speeds of 55 miles per hour.
While the findings indicate AV robotaxis are a considerably safer alternative to traditional cabs, it should be noted that Cruise and its parent company GM are the backers of the study.
Currently, a series of high-profile traffic snarls and a flurry of complaints continue to follow Cruise AVs wherever they are deployed. A swarm of Cruise AVs caused a traffic jam in Austin, Texas recently, unable to maneuver around each other “even though they’re all part of the same network” according to a witness. The snarl, called a “crowding event” by the company, was eventually resolved remotely.
Cruise has also been bombarded with complaints in the Texas city, ranging from simply blocking the way to occasional accidents, though most of the latter consist of a human-driven vehicle ramming a Cruise AV rather than vice versa.
Notably, however, and in keeping with the study’s findings, the recent incidents involving Cruise AVs have snarled traffic, but resulted in no fatalities or injuries.
Cruise CEO Kyle Vogt says the incidents are sensationalized by the media, summing up his position by saying “we’re talking about a 15-minute traffic delay for something that, on the other hand, is providing a massive and quite measurable public benefit to the community.”
Meanwhile, Cruise is awaiting the regulatory green light from the NHTSA to begin production of the Cruise Origin, a vehicle built from the ground-up as a dedicated self-driving robotaxi.
Subscribe to GM Authority for more GM Cruise news, GM technology news, and around-the-clock GM news coverage.
Comments
A link to the study would be helpful.
Establishing a Crash Rate Benchmark Using Large-Scale Naturalistic Human Ridehail Data FINAL.docx
Bunch of self propagating BS. The “sample size”,of the mileage is a joke since humans drive about 6 million miles a minute in this nation and Cruise is geofenced. I’m sure U of M gets some nice financial support from gm.
You have a problem with sample size? Well, you have a “sample” of 100% for Cruise, which most statisticians would regard as an adequate “sample”, and there is plenty of data available for taxis.
And the University of Michigan is not a slouch outfit – if they did the study, it’s a good study.
As said in another comment, a link to the study would be useful.
I have found that figures can be projected in a manner to either support a topic, or they can be projected to kill a project.
I agree that this news release is less than accurate and is actually misleading for self promotion.
If there are people that really believe robot vehicles are safe, please contact me.
I will sell you acres of swamp land for development, that I don’t even own !
Statistics has often been defined ad “the ethics of numbers”
I believe the phrase you are looking for is “There are three kinds of lies: Lies, Damned Lies, and Statistics” sometimes attributed to Twain.
Yes, data can be manipulated. It helps if the data is from a trusted source. So you’re saying you don’t trust the University of Michigan. So who exactly would you trust to do a study like this? A major university with experience in the transportation industry, or, let’s say, Fox News?
A major Liberal university with plenty of funding provided by gm among others.
As always, Cruise ignores the fact their vehicles are geo-fenced, have restricted hours of driving, and not allowed to drive in adverse weather conditions, not even fog or simple heavy rain.
Cruise, show us how you only pulled human taxi-driver data for your geo-fenced locations, restricted driving hours, and perfect weather.
Totally meaningless.
This is a no brainer. Thre are more human driven taxi’s than there are Cruise’s. So, it would stand to reason that the ratio would be lower. A third-grade math student (except Baltimore’s) could figure this out.
You clearly don’t understand statistics and sampling. 100% sample of Cruise, a statistically significant sample of taxis. There is no sample bias. Anyone within shouting distance of an academic institution will not produce a study with sample bias – they would lose ALL credibility in an academic institution/community.
Cruise commissions a study and surprise, cruise robots are safer. News at 11.
Commissioned from the University of Michigan – there is no organization better qualified to conduct such a study than
Happy to see that you made the correction from “less” to “fewer.”
Commissioned from the University of Michigan – there is no organization better qualified to conduct such a study.
OK, let’s leave it this way – if someone is killed or injured by a Cruise vehicle, and the case goes to court, all of these statistics will be closely examined by experts, and the truth of the matter will be legally established.
To my knowledge, Cruise has not been shown to be unreasonably unsafe. Not perfect, but human drivers aren’t perfect, either.
Some really persnickety California regulators examined the data and found Cruise to be reasonably safe. And GM collected whatever data the regulators wanted, so it’s all there. Cruise has the data to show it’s safe.
GM has data, certified by a government agency/regulators. What data to you have? By the way, an opinion based on no data does not qualify as data….
It’s important to note that it’s not taxis but rather “ride hail” drivers who tend to be younger and male – – the cohort which is involved in more accidents. Taxi drivers are usually older, plus they’re professional drivers.