From a user-centric perspective, there have been huge risks in leaving ADAS and AV safety guidelines entirely up to industry stakeholders or "lower" government agencies.  In this status quo, decision-making rarely starts from a viewpoint of the general public's safety.  Rather, industry stakeholders (who must answer to boards or investors) are more interested in profit goals, which require racking up mileage and clout.  And in the U.S. especially, various non-federal agencies (largely unknowledgeable on current tech, but eager to be seen as tech-friendly to encourage job growth) are creating an uneven patchwork of industry-friendly regulations.

With this as backdrop in the U.S., late last week (ironically after the NTSB scolded NHTSA, Uber, and the state of Arizona for their roles in a deadly crash), NHTSA publically requested comments on its first test procedures for various ADAS features.  Methodologies have been proposed for nine different features (including advanced park assist and traffic jam assist), and a public comment deadline is set for January 2020.

In these uncertain times where ADAS/AV stakeholders have dominated the discourse with very public on-road tests and related charm offensives, NHTSA is to be commended for finally contributing some type of leadership, even if it is imperfect.  But as with many other NHTSA efforts (like their mobile device UI guidance released in 2016), these guidelines are long past overdue, and very imperfect.  Worst of all from a human-centric perspective, these guidelines seem to have been crafted with the goal of merely streamlining the testing process, rather than creating a truly useful process which improves safety for all road users.

Proposed tests for park assist and traffic jam assist (features Strategy Analytics is quite familiar with through our series of evaluations, most recently with Chrysler ParkSense and Nissan ProPILOT Assist) provide a number of troubling illustrations.

• Testing guidelines require a dry and flat road surface, good weather conditions, good visibility, and without facing the sun during low-sun-angle time periods.In other words, system safety is only to be tested in the best or most favorable conditions for cameras, radars, and other system components.
• Parking space detection is only to be tested using the right-hand side of a vehicle.No rationale for this limitation is given.
• The lead vehicle or obstacle to be used during testing must only be a white compact car.This test parameter does nothing to address known safety issues, given the problems features like Tesla Autopilot have had in detecting certain objects like emergency vehicles or concrete barriers.

• The park assist protocol makes numerous recommendations with regard to operator HMI, but implicitly encourages text-heavy alerts, even at timepoints requiring the operator to monitor their surroundings.And while the park assist protocol takes steps to outline what they think a safe HMI might look like, the traffic jam assist protocol has no similar focus.

Safety for car owners, pedestrians, and other roadway occupants cannot just be a noble goal, it must be standard practice.  All ADAS features have a role to play in this practice, and thorough testing of these features (with a mission of improving public safety) must be a part of that practice as well.  If the automotive industry (or nontraditional stakeholders such as Waymo or Uber) cannot prove that they can be transparent or trusted partners in public safety, government agencies are likely the best way to achieve this.  But thus far, government agencies seem incapable.  In fact, NHTSA's very belated announcement highlights the same issue we noted when NHTSA released its mobile device UI guidelines:  Regulations as they are now simply cannot keep up with technology, and will be irrelevant by the time they are actually implemented.

Regardless of where safety initiatives originate, they must be proactive rather than reactive to technological advancements.  The fact that U.S. agencies are just now, in 2019, attempting to draft very basic, streamlined, industry-friendly protocols for ADAS testing is commendable, but sad.  It is much like when Uber announced the creation of a “safety advisory board” a full 18 months after one of its test vehicles killed Elaine Herzberg.  Both moves are pointless reactivity that will do nothing of true substance to keep the traveling public safe.