Featured Image Source:  Reuters (A Neolix self-driving vehicle, seen at the IEEV New Energy Vehicles Exhibition in Beijing)

Background

Safety performance will be the determinant factor in not only the long-term viability of automated driving system (ADS) vehicles, but the short-term perception and branding of the “Future of Mobility” to the general public.  The actual promise of the future of mobility is as diverse as e-bikes, e-scooters, robotaxis, Mobility as a Service (MaaS), supersonic jets, and air taxis. For now, however, the broad category of public and personal mobility and autonomy and autonomous technology systems are tethered to the overhyped promise of self-driving cars.

The techno-utopian marketing and branding have been so fierce about this subsector, that the general public (consciously or subconsciously) maps their tracking of the issues solely to the future of the automobile.  Such a simplification is analogous to equating weekend Hollywood Box Office numbers to any of the metrics one is able to garner from the video streaming platforms (Hulu, Discovery+, etc.) – which, ironically, the general public may also do.

Quantitative data and performance metrics will matter- and the vital metric to track in the ADS vehicle subsector is safety performance and, at a more granular level, collision and fatality measures and rates.

The Latest ADS Collision Data from the NHTSA

Our source for all things autonomous vehicles is “Egil’s Eye” – which is the regular column of Egil Juliussen over at the EE|Times.

Juliussen has over 35 years’ experience in the high-tech and automotive industries. Most recently he was director of research at the automotive technology group of IHS Markit. His latest research was focused on autonomous vehicles and mobility-as-a-service. He was co-founder of Telematics Research Group, which was acquired by iSuppli (IHS acquired iSuppli in 2010); before that he co-founded Future Computing and Computer Industry Almanac. Previously, Dr. Juliussen was with Texas Instruments where he was a strategic and product planner for microprocessors and PCs. He is the author of over 700 papers, reports, and conference presentations. He received B.S., M.S., and Ph.D. degrees in electrical engineering from Purdue University, and is a member of SAE and IEEE. (1) 

In his June 22nd column, Latest ADS Collision Data from the NHTSA, Juliussen summarizes data from National Highway Traffic Safety Administration (NHTSA) documents released in June and “additional collision data.”  The  NHTSA dataset covers July 2021 through May 15, 2022.

The NHTSA plans to release similar data on a monthly basis. You can download several datasets directly from the NHTSA below:

• NHTSA press release
• ADS and ADAS summary
• ADS summary PDF file
• ADS detailed Excel data by incident
• Other incident Excel data
• Data element definitions

The Taxonomy: The Six Levels of Driving Automation

Any discussion of ADS vehicles (and any future discussion in these pages of Advanced Driver Assistance Systems  – ADAS 2), is framed by the “Levels of Automation” taxonomy initially provided by SAE International and adopted by the NHTSA:

“The Society of Automotive Engineers (SAE) has published a document describing a taxonomy to classify systems into Six Levels of Driving Automation. These SAE Levels are used to describe varying degrees of driving automation that perform all or part of the dynamic driving task on a sustained basis and vary from Level 0 (no driving automation) to Level 5 (full driving automation).”

Source: SAE International

Source:  Automated Vehicle Safety | NHTSA

J. D. Power also has an accessible breakdown of the taxonomy here.

About the Report: the NHTSA Standing General Order on Crash Reporting

According to the June report:  “NHTSA has issued a Standing General Order (the General Order) requiring identified manufacturers and operators to report to the agency certain crashes involving vehicles equipped with automated driving systems or SAE Level 2 advanced driver assistance systems. The General Order allows NHTSA to obtain timely and transparent notification of real-world crashes associated with ADS and Level 2 ADAS vehicles from manufacturers and operators. With these data, NHTSA can respond to crashes that raise safety concerns about ADS and Level 2 ADAS.”

Following is a review of the data by Egil Juliussen:

“NHTSA ADS Data: The NHTSA released three sets of data. The ADS, or L3 (Level 3:  Driving Automation – Conditional Driving Automation) – L4 (Level 4:  Driving Automation – High Driving Automation), data includes summary information in a PDF file and detailed incident information in an Excel file. There are 130 incidents from July 2021 to May 15, 2022. ADAS, or L2 (Level 2: Driving Automation – Partial Driving Automation), information is the second dataset and has 367 incidents.

The third, or “Other,” dataset has a large number of incidents, but few have detailed information.

The NHTSA provided an eight-page summary of the ADS data…The data covers 130 collisions from July 2021 to May 15, 2022.

There is also an Excel-ready file with detailed information on all collisions. The file has 207 entries and covers 130 crashes. The extra entries are due to multiple fillings per incident, including updates with new information. Each spreadsheet entry has detailed collision information — up to 122 columns of data. The Data Element Definition file explains what each column means (linked above). Useful information before the crash, during the crash, and after the crash is requested, including vehicle movements, weather and lighting conditions, road conditions, location coordinates, addresses, and more. Personal and related information are redacted.

[This] table summarizes the names of the companies that filed crash reports. The data is from a figure on page 5 in the NHTSA summary report. For better perspectives, the information is segmented into ADS use cases — robotaxis, MaaS, and robotrucks.

There is also a segment for all the auto OEMs. It is possible that Ford data is a duplicate of Argo.ai data and GM data is a duplicate of Cruise data. The ‘Other’ category lists one Tesla filling that should have been in the ADAS L2 category. Due to multiple fillings for some crashes, the total is 191 for the 130 crashes. Local Motors Ollie has two duplicate fillings by Beep and Robotic Research:

• The robotaxi category has the most crash listings at 113, with Waymo leading with 62 crash reports.
• MaaS has 43 crashes due to a large number by Transdev.
• Robotrucks has only two items.
• However, the NHTSA’s ‘other incident reports’ had a large number of truck crash reports.

The [above] figure shows injury severity for ADS crashes, and the data is promising. The figure is from page 6 of the NHTSA’s ADS summary report:

• Of the 132 crashes reported, 108 (82%) had no injuries.
• Only 12 crashes (9.1%) had minor injuries.
• Moderate injuries occurred in three crashes (2.3%).
• One crash had a serious injury, which is less than 0.8% of total crashes.
• There were no fatality crashes.

The NHTSA has satisfactory data on exactly what ADS vehicles collide with, as shown in the next figure. The data is from page 6 of the NHTSA’s ADS summary report.

• Passenger cars are the leading category, at 61 (46%) of the total 132 items included.
• SUVs accounted for 20% (27) items.

State crash data:  The ADS crash data is available by state. California accounted for most of the data, at 90 of the 130 crashes (69%). Arizona was next, with 12 crashes (9.2%). Florida was third with seven crashes (5.3%). Nevada had four crashes. Five states had two crashes: Colorado, Indiana, Michigan, Virginia, and Wyoming. Two states had one crash: Georgia and Pennsylvania.

ADS crash report sources:  Most of the crashes were reported by multiple sources, as there were a total of 258 crash report sources listed for 130 crashes:

• Telematics was the leading source of crash reports, with 104 reports (80%) of all crashes. Field reports had 78 items (60% of all crashes).
• Testing of ADS vehicles provided 62 reports, or nearly 48% of all crashes.
• Complaints or claims added four crash reports.
• Law enforcement added only one crash report. Eight reports were unclassified.

NHTSA other data

The NHTSA provided an Excel file with other incident data. It is the largest file, with 912 collision entries including duplicates. Analysis estimates there are 21 duplicates. The file has the same data structure as the ADS and ADAS files, but there are very little details available for most entries.

The problem with this data is that both ADS and ADAS vehicles are included and there is little data to split the two segments. To get some clarity on the ADS vehicles, this analysis uses the same company segmentation used for ADS entities listed above. This means the companies listed in robotaxis, MaaS, and robotrucks will be in the ADS category. Most of the “Other” category are likely ADS vehicles.

The auto OEMs contain both ADS and ADAS L2 vehicles, with ADAS likely accounting for the bulk of collisions. Luxury auto brands such as Bentley, Ferrari, Lamborghini, Maserati, and McLaren were excluded and they are mostly ADAS L2 vehicles.

Also excluded were Tier 1 companies such as Continental, Denso, Magna, Qualcomm, Robert Bosch, Valeo, and ZF. They are also likely ADAS L2 vehicles.

The below table includes 646 collisions from 74 companies. The robotaxi, MaaS, and robotruck categories add up to 359 collisions, and they should all be ADS vehicles. About 70% of the other category are likely ADS vehicles. The percent of ADS collisions of the auto OEM category is in the 30% range. This means about 500 of the collisions in this NHTSA database are ADS incidents.”

What Next?

• “It is helpful to get NHTSA collision data for ADS vehicles, but much more is needed. We also need vehicles miles traveled to get better analysis of ADS improvement trends.
• There is also collision data from the California DMV. A recent column looked at the ADS collisions tracked by the California DMV and is available here.
• We also need to have a better understanding of just how good ADS vehicles must be. One would expect that they should perform better than the average driver in terms of collision rates.
• The average miles between fatalities decreased from 9.037 million miles in 2019 to 7.48 million miles in 2020, or 17.2%.
• NHTSA prefers to use a different fatality measure, expressed as fatalities per 100 million VMT. In 2019, the fatality rate was 1.11 fatalities per 100 million VMT, which jumped to 1.34 fatalities per 100 million VMT in 2020.
• The fatality rate stayed about the same in 2021 at 1.33, as 42,915 people were killed in 2021 in the U.S. This was due to much higher VMT in 2021 at about 3.23 billion miles, or an 11% increase from 2020.”

The long view?  The adoption rate by consumers of ADS vehicles will be a roller coaster ride in the next five to eight years based on these metrics  – and the general public’s perception of the safety and viability of a future “driven” by the self-driving car market.