//php echo do_shortcode(‘[responsivevoice_button voice=”US English Male” buttontext=”Listen to Post”]’) ?>
Waymo and Cruise are busy testing their robotaxi software in San Francisco. These two companies are among the leaders in autonomous vehicle (AV) investments. My last column compared their efforts based on the public data available from the California Department of Motor Vehicles (DMV). This column summarizes data from the California Public Utility Commission (PUC) and California DMV collision report.
The California DMV requires AV companies to file reports when an AV is involved in a collision, and this includes everything — even a slight bumper incident. The data is available at the California DMV’s website.
Each collision report is a downloadable three–page PDF file. Each report includes a lot of data such as date, time, location, weather, lighting, and road conditions. A section on injuries/death and property damage is included. Details on the collision include movement before the collision, type of collision (head–on, sideswipe, etc.), roadway characteristics, AV damage, and information on other parties involved in the collision.
A one or two paragraph description of the collision history and outcome was the most useful part of each collision report.
The next table is a summary of the yearly collision data for 2015 through 2021 and a sum for the seven–year time period. Total collisions for all companies are included in the table along with Waymo’s and Cruise’s data. Waymo and Cruise’s share of total collisions are listed. The yearly AV miles driven adds perspectives on the yearly collision trends.
Waymo’s collision trends show improvements by year. As Waymo began testing in San Francisco in 2021, the number of collisions increased along with much higher AV miles travelled. The Cruise collision trend versus AV miles travelled also show improvements starting in 2018.
The number of collisions per million miles can be calculated from this data, but since most of the collisions are caused by other vehicles it may not be very useful.
The reports do not list who caused the collision, but there is enough data to make an assessment in the vast majority of incidents. It takes a lot of effort to compile and analyze the data, however.
To obtain some perspective, I looked at Waymo and Cruise collision reports for 2019, 2020, and 2021. The California DMV website has collision data for 2022 through early May 2022, which was also included. The next table is a summary based on my assessment of the collision faults from analyzing the collision reports — about 260 between Waymo and Cruise.
The table shows that the vast majority of the collisions are caused by other vehicles and road users. For Waymo, 79% of the AV collisions were the fault of other road users. It was even higher for Cruise with 87% of collisions caused by other road users.
Waymo had two collisions that looked like they were caused by the AV system—both were bumping into parked cars in a multi-point turnaround. Cruise had one collision that looked like it was the AV’s fault.
The AV safety drivers were at fault in 27 collisions for Waymo, which is 19% of Waymo collision reports. Cruise safety drivers were at fault in 14 cases, or 13% of Cruise collisions.
It is speculative to make any conclusions from this collision data — there simply is not enough data yet. For perspectives we can look at U.S. crash rates using 2019 data since it was a “normal” year for crash statistics. The U.S. crash rates increased in 2020 and 2021 due to aggressive and speedy drivers.
The NHTSA Traffic Safety Fact publication for 2019 shows the data below. NHTSA also calculated crash rates per 100 million vehicle miles travelled (VMT).
- 2019 VMT: 3.262 billion
- 2019 Total police reported crashes: 6,756 ⇒ Crashes per 100M VMT=208
- 2019 Fatality crashes: 32,244 (36,096 total fatalities) ⇒ Fatalities per 100M VMT=1.10
- 2019 Injury crashes: 1,916 (2,740K total injuries) ⇒ Injuries per 100M VMT=84
- 2019 Property–only crashes: 4,806 ⇒ Property crashes per 100M VMT=147
In 2020 and 2021, the U.S. fatality rate grew to 1.34 and 1.33 per 100M VMT, respectively.
California PUC Data
The characteristics of the AV miles in the quarterly data collected by the California PUC is the key for getting the permits for robotaxi service in California. Cruise and Waymo are the leaders and are likely to be the first to deploy such services in California, with San Francisco as the first area.
A previous column has California PUC data through November 2021 for other companies.
Eight companies have received drivered AV permits from PUC: Argo.ai, AutoX, Cruise, DeepRoute.ai, Pony.ai, Voyage Auto, Waymo, and Zoox. The permits for Pony.ai and Voyage have been suspended. Cruise is the only company that has received a driverless permit from the California PUC.
The California PUC is collecting AV usage data to track the use of electric vehicles and the efficiency of AV operations in terms distance before passenger pickup and the vehicle’s waiting time for passengers. The next table summarizes Waymo and Cruise data and the total for all companies. The 2022 data is only for January and February. The California PUC will report March–May data in early July 2022.
The California PUC data shows that Waymo has used the most AVs and reached 301 drivered AVs in February. Cruise peaked at 69 drivered AVs in 2021 and is switching to driverless AVs in 2022 with 19 in February 2022. It is likely that Waymo will receive a driverless AV permit in the near future.
Total AVs used by all permit holders were 104 in 2019 and 107 in 2020. This grew to 337 in 2021. Total AVs in 2022 was 320 in February 2022 but is likely to grow as other companies restart their testing after the pandemic.
Total miles driven
Regarding total AV miles, Waymo is again the overwhelming leader with over 2.7 million drivered miles from 2019 through February 2022. This gives Waymo over 95% of the drivered miles in the California PUC data. In 2019 and 2020, Waymo was driving in Palo Alto-Mountain View areas. In 2021, Waymo’s driving shifted to San Francisco.
Cruise has been driving in San Francisco starting in 2020. Its drivered milage is only 915 miles as it shifted to a driverless permit in November 2021. Total driverless AV miles were 3.8 thousand at end of February. Cruise is the only company with a driverless permit from the California PUC.
Since California is the U.S. leader in EV use, it is no surprise they want to have statistics on robotaxi EV usage. For all companies from 2019 to 2022, EV miles accounted for 70.1% or over 2.01 million miles. Waymo did not use EVs in 2019 and 2020 but was at 99% in 2021 and 2022. Cruise only used EVs in all of AV testing — drivered and driverless.
All robotaxi companies want minimal idling time since it is one parameter that improves operational efficiency. Waymo accounted for 87.9% of all idling time or over 79 thousand hours.
The California PUC data has no measure for total hours of robotaxi operation. This makes it difficult to assess idling time. If we assume that the average speed of Waymo’s AVs was 15 mph, the driving time for Waymo’s 2.7 million miles would be 180 thousand hours. This gives an idling time plus driving time of nearly 260 thousand hours. With this average speed guess, idling time for Waymo is then 30% of total operational time, which seems on the high side.
AV miles to passenger pickup
This is another important robotaxi operational parameter. For all AV companies, the AV miles before passenger pickup was 133,540 miles, which is 4.7% of total miles. Waymo had 131,411 miles before passenger pickup or 4.86% of Waymo’s total miles from 2019 through February 2022.
Total passenger transport for all AV companies was nearly 85,476 from 2019 to February 2022. Waymo was dominant with 68,480 passengers or 80% of total.
Cruise carried 1,844 passengers including passengers from both drivered and driverless AVs. The driverless AVs accounted for over 75% of Cruise’s total passengers.
This column compares Waymo and Cruise’s AV testing in California as they are leading the parade to future robotaxi services with San Francisco being the likely first city to realize such services.
The California DMV requires reports of any collisions where AVs are involved. From 2015 through 2021, a total of 393 collision reports were filed for all companies that tested AVs in these seven years. Waymo reported 149 collisions while Cruise reported 167. The vast majority of these collision reports are caused by other vehicles or road users — not by the AVs.
A more detailed analysis of 260+ collision reports for 2019-2021 and part of 2022 was completed for Waymo and Cruise. By reading and analyzing these reports, the fault was assessed for each collision. Based on this analysis, it looked like the AVs could be blamed to have caused the collision in three of the 161 collision reports. A Waymo AV operating in AV mode accounted for two of the collisions while Cruise was at fault in one collision.
The AV safety drivers were at fault in 41 collisions — 27 for Waymo and 14 for Cruise. The remaining 217 collisions were caused by other vehicles and road users, which is 83% of total incidents.
It is difficult to draw conclusions from the collision data and easier to pose questions. What does it mean that Waymo and Cruise were at fault in an estimated 3 of 261 collision report? Can we calculate meaningful collision data per 100 million miles for Waymo and Cruise’s total AV miles? We need more data to answer these and other questions.
Waymo has been the dominant participant in the California PUC AV testing in terms of AVs used, miles driven, EV miles, and passengers. Waymo accounted for over 95% of AV miles driven, nearly 94% of EV miles, and over 80% of passengers transported. The next step for Waymo is to receive a driverless permit from the California PUC.
Cruise is the leader in driverless AV testing since it is the only company with a PUC driverless license. The next step for Cruise was to receive a PUC permit to charge customers for robotaxi services, which occurred on June 2, 2022.