On Wednesday, February 4th, the Senate Committee on Commerce, Science, and Transportation held a hearing on Autonomous Vehicles (AVs), titled “Hit the Road, Mac: The Future of Self-Driving Cars.” Leading up to the hearing, the League of American Bicyclists submitted a letter for the record, reiterating our support for an AV regulatory framework that includes a “vision test” ensuring safety for people biking and walking and incorporates the safety tenets we developed with other safety organizations.
Watching the roughly two and a half hours hearing, I don’t believe that either a “vision test” or the “tenets” were mentioned. But, two key phrases emerged in testimony and questions that I think are likely to end up as important parts of any federal regulatory framework that is put forward under this Congress.
Before turning to some technical discussion of those key terms, there were a few standout remarks by Senators and witnesses that deserve highlighting for those who don’t want to read through my 60ish Bluesky posts.
- Senator Maria Cantwell, the committee’s ranking member from Washington State, made a great comment at the opening of the hearing that whatever the promise of AVs, we have technology that can save lives today like Automatic Emergency Braking (AEB). We are currently asking people to contact Congress expressing their support for cyclist-AEB so that people who bike will get the safety benefits of AEB and Senator Cantwell’s comment highlighted this currently available technology.
- Not many senators mentioned specific legislation, but two mentioned bills the League has supported:
- Senator Ben Ray Luján (D-NM) talked about the importance of the HALT Act, which requires technology to detect and prevent drunk driving. The League supported the inclusion of the HALT Act in the last transportation bill and looks forward to its implementation.
- Senator Ed Markey (D-Mass.) talked about the Stay In Your Lane Act, which requires AV companies to report on their Operational Design Domains (ODD) and not operate outside of ODD for which they are designed. ODD are used by AV companies to ensure that systems can handle different contexts, such as dense urban areas, school zones, or rural country roads. The League has signed on in support of the Stay In Your Lane Act and hopes that it is incorporated into the next transportation bill.
- Witness Bryant Walker Smith, a law professor at the University of South Carolina, made a point that Senator Cantwell reiterated: “people are dying today not because we are careful about automated driving but, rather, because we are careless about road safety generally.” While he is an expert on AV law, he noted that the difference in traffic safety between the United States and other countries is not “a vast secret fleet of automated vehicles” but instead the ability to safely “cross the street” and encouraged senators to think beyond AVs to improve traffic safety.
Now onto the two key terms likely to determine whether our priorities are included in an AV regulatory framework.
Safety Case
Several senators asked about different aspects of what a safety case is and how it might address questions about AV safety from their constituents. Both Waymo and the Autonomous Vehicle Industry Association made requiring a safety case a core part of their appeal for a federal regulatory framework. According to the written testimony of Waymo, a safety case is “a structured argument, supported by a body of evidence that provides a compelling, comprehensible and valid case that a system is, or will be, adequately safe for a given application in a given environment.”
A safety case requirement can be consistent with our priority of a “vision test,” but a safety case is also different from how the National Highway Traffic Safety Administration (NHTSA) has typically approached vehicle regulation. A safety case can often lack performance data or specific performance standards. Most of NHTSA’s vehicle regulation that currently exists is through Federal Motor Vehicle Safety Standards that specify the performance of vehicle components and manufacturers self-certify that their vehicles meet those performance standards. The worst case for a safety case is that it is self-certification on steroids, essentially serving as documentation of internal processes, and is required but not public.
Behavioral Competencies
As the League has advocated for a vision test for nearly a decade, we have described it as a test that verifies minimum performance standards related to how AVs detect and respond to all road users, including people biking and walking.
Behavioral competencies may capture much of what we want from a “vision test” for AVs. Discussion of behavioral competencies primarily came from the Autonomous Vehicle Industry Association, which listed it as one of the key ways that senators could take action to promote American leadership in AV technology. While it is not a term that occurs frequently in policy documents issued by USDOT, it is a term with a history of development through California regulations, research, AV industry participants, and several best practice documents issued by the Automated Vehicle Safety Consortium (AVSC) program of the SAE (formerly Society of Automotive Engineers) Industry Technology Consortia.
Behavioral competencies are often described in terms similar to what we would expect from a vision test and behaviors related to people biking and walking are included. In the AVSC document, “Best Practice for Evaluation of Behavioral Competencies for Automated Driving System Dedicated Vehicles,” they describe behavioral competencies as an “approach, to some extent, [that] resembles a human driver test.” Within the Elemental Set of Behavioral Competencies provided in that AVSC best practice document, “Responding to Vulnerable Road Users” by “maintaining a safety envelope with respect to VRUs” is one of 13 behaviors listed. Similarly, in a California “Peer Review of Behavioral Competencies for AVs” published in 2016, “detect and respond to bicyclists, pedestrians, and animals” is one of ten behavioral competencies reviewed.
The House of Representatives includes behavioral competencies as “ADS (Automated Driving System) competencies” in the latest version of the SELF DRIVE Act, specifying that one competency required is that “[a]n ADS can detect and respond appropriately to any vulnerable road user likely to be present and in proximity to the ADS in the relevant ODD (Operational Design Domain).” In that legislation, the safety case it requires would include “[a]n explanation of how the ADS-equipped vehicle meets each competency listed [in the paragraph describing ADS competencies].”
In order to be similar to a “vision test” as promoted by the League, behavioral competencies would have to be actually shown through data. Mere explanation as contemplated by the safety case in the House’s SELF DRIVE Act is not enough. Actual evidence of behavioral competencies through objective data that shows the safety envelope afforded in several specific scenarios or other important context-sensitive behavior, such as slowing down, is what we mean when we say “vision test.”