Mixed Messages on MaaS Market Readiness
Data source – California DMV, presented by IDTechEx

IDTechEx’s Analysis of New Driverless Vehicle Testing Data From California DMV

Author: Dr James Jeffs, Senior Technology Analyst at IDTechEx

Over the last three to four years, the driverless robotaxi industry has begun to flourish. Driverless services are coming online in multiple cities across the US and China. IDTechEx’s recent report, “Future Automotive Technologies 2024-2034: Applications, Megatrends, Forecasts”, predicts that the driverless robotaxi industry will be generating over US$470 billion annually through services in 2034. This is certainly supported by the driverless data coming out of California DMV, which shows another year of more miles driven, by more vehicles on the road, and more distance between disengagements. However, there are two sides to every story, and the news coming out of San Francisco suggests that not everything is getting better.

What the autonomous testing data says

In 2023, the fleet of vehicles testing in California submitted more testing miles than ever before. Alphabet’s Waymo has always been a leader in the number of miles driven during testing, and this year was no different. In 2023, Waymo submitted nearly 3.7 million miles using its fleet of 438 registered vehicles, which is around 65% of all autonomous vehicle testing miles submitted in 2023. This is the number of miles submitted with a safety driver onboard testing the systems and making sure everything is working as expected. So perhaps more important is the number of miles submitted through vehicles without a driver on board at all. In 2023, Waymo submitted a staggering 1.2 million miles of driverless testing, more than all miles recorded by all testing companies in all years prior. However, this was beaten by another of the leading players in the robotaxi race – Cruise.

Cruise operated a commercial robotaxi service in San Francisco for most of 2023. During its testing, it racked up over 2 million miles of driverless activity, nearly double that of its main rival, Waymo. However, what was interesting is that Cruise did not put anywhere near as many miles into its safety-driver-monitored autonomous testing. Despite having a fleet of over 500 vehicles registered for driver-in autonomous testing, only 325 captured any miles, and they only clocked a total of 580,000 miles.

Where Cruise did beat Waymo was in its disengagements. Disengagements happen when the test driver feels the vehicle is not behaving as planned. This can be because the sensor suite has failed to detect something, the planned trajectory of the vehicle is considered dangerous, or even because a human motorist nearby is driving erratically (which is surprisingly common in the data). Last year, Waymo submitted 212 disengagements, giving them a ratio of 17,311 miles per disengagement, which isn’t great. For comparison, the average US human driver can expect to be involved in a collision roughly once every 200,000, according to IDTechEx’s research. Cruise, on the other hand, recorded no disengagements - zero, zilch, nada! Cruise’s disengagement performance has been increasing rapidly over the years; in 2022, it only had nine disengagements, four of which were caused by poor driving by humans in nearby cars. Even still, to record zero disengagements over 580,000 miles of testing is quite the achievement.

According to the data, the robotaxi industry is getting more experienced and safer with each passing year. But the streets of San Francisco are telling a different story.

What the news and events in San Francisco say

There has always been some controversy over autonomous car testing in San Francisco and across California. They have seen pushback from taxi drivers and ride-share drivers, with the new technology posing a threat to their livelihoods. However, the biggest worry when it comes to the deployment of driverless autonomous services is their safety.

Over the course of robotaxi testing in California, there have been several small and perhaps even silly incidents involving these vehicles. For example, on the 21st March 2023, a Cruise vehicle went through the caution tape of a closed intersection where there was a downed telegraph pole, which the Cruise vehicle then hit. But, on the 2nd of October 2023, an incident sparked a deeper sense of concern. A pedestrian was hit by a human-driven Nissan, which launched her into the path of a Cruise vehicle. The Cruise vehicle failed to stop in time and, while attempting to pull over, dragged the pedestrian 20ft before stopping with a wheel on her leg, causing serious injury. Following the incident, Cruise’s driver-out testing and robotaxi service permissions were revoked by California DMV. Rivals Waymo were still able to operate but has recently had challenges of its own.

In Feb 2024, a Waymo vehicle had a collision with a cyclist, causing minor injuries. This happened as the cyclist was obstructed from the Waymo’s view by a truck. The cyclist emerged into the path of the Waymo, which braked heavily as soon as the cyclist became visible but was unable to avoid a collision. The cyclist suffered minor injuries but didn’t need to be taken to hospital and was able to leave the scene on their own.

Later in the same month, a Waymo vehicle was set on fire and vandalized after it entered the crowded streets of San Francisco’s Chinatown on the busy night of the Lunar New Year. Some have claimed that this was a mistake on behalf of the car and that a human driver would know to avoid the area during this time. However, this should not excuse the actions of the crowd.

Of course, since testing began, there have been collisions involving robotaxis. California DMV has a log of all collisions that involve autonomous testing vehicles, a log that IDTechEx has read and analyzed. Between 2019 and 2023, nearly 550 collisions have been recorded; of those, 288 occurred when the vehicle was in autonomous driving mode, as opposed to in manual mode being driven by a test driver. Of those 288 collisions, IDTechEx finds that 22 can be attributed to a fault of the autonomous driving system. Furthermore, only 11 occurred when the vehicle was operating without a safety driver onboard, and only one caused a major injury.

There have been other injuries involving autonomous vehicles, but aside from the incident on October 2nd, only one other minor injury has occurred when the vehicle was without a safety driver; that one was ultimately down to an unfortunate tire blowout. So, there has been one major injury caused by an autonomous robotaxi in nearly 4 million miles of driver out testing. By comparison, NHTSA reported a total of 2,497,647 injuries from road traffic accidents in the US in 2021. At the same time, the total fleet of vehicles on the road completed just over 3.1 trillion miles. In other words, the typical injury rate on US roads is once every 1.25 million miles.

While what happened on October 2nd was undoubtedly a terrible incident for the person involved, the data does indicate that robotaxis are already causing fewer incidents than human drivers. Additionally, the silver lining of this tragic collision is that Cruise can learn from it, update its software, and update its fleet. If the same incident were to be repeated, it should perform much better. Human drivers do not do this. Humans have the same collisions all the time, and while one driver might learn and improve, it will have little to no impact on the rest of the drivers. For these reasons, the industry needs to keep pushing ahead with autonomous technologies.

IDTechEx has tracked the performance of robotaxis since records began and uses this information to inform its autonomous vehicles forecasts, which can be found in key industry reports “Autonomous Cars, Robotaxis and Sensors 2024-2044” and “Future Automotive Technologies 2024-2034: Applications, Megatrends, Forecasts”. Downloadable sample pages are available for all IDTechEx reports.

For the full portfolio of autonomous vehicle market research from IDTechEx, please visit www.IDTechEx.com/Research/Autonomy.