Product1 publisher3 min readPublished
NHTSA and SAE launch ASCEND to write the driving test for automated vehicles
The question of when a robot driver is good enough has been the industry's to answer. NHTSA's new consortium with SAE is meant to produce the competency test that would settle it, with the agency's rulemaking to follow.
The Product Desk · Product desk
What happened
- Federal motor vehicle safety standards were written around required equipment such as pedals, steering wheels and mirrors, none of which the Tesla Cybercab and Zoox robotaxi have.
- NHTSA has reached an agreement with SAE to launch ASCEND, a government-industry initiative meant to speed up the development of data-driven AV performance standards.
- No timeline has been set for when the consortium will actually produce automated driving system standards.
Compiled by The Product DeskSomething wrong?How this is made
Why it matters
- constraint Until a competency standard exists, a city or an insurer sizing up an operator can only check the metrics that operator chose to publish, on the routes it chose to run.
- decision Whatever ASCEND settles on as competency decides which telemetry counts, so teams building logging and evidence pipelines this year are guessing at the contents of the test.
- precedent Morrison points the standard at state and local requirements. The consortium's metric choices are likely to be the single bar most operators have to clear.
- exposure A pure-play robotaxi listing hands the unsettled safety question to public-market investors, who would price it with no outside performance test to reference.
Ask a robotaxi operator whether its software is a safe driver and you get the operator's answer, on the operator's metric. TechCrunch's Mobility newsletter reports that the question of when the technology is safe enough has largely fallen to industry to answer [2], and that no standardized test exists for judging how an automated driving system performs in the real world [1]. There is no exam for the software, the way there is for a person before licensing [3].
The gap has two parts. The first is equipment: federal motor vehicle safety standards were built around pedals, steering wheels and mirrors, and the Tesla Cybercab and Zoox robotaxi do not have them [5]. The second is competency, meaning whether the driving is any good. ASCEND, the government-industry initiative NHTSA agreed with SAE to launch, aims at the second and is pitched as a way to speed up data-driven performance standards [8]. "Our top priority for this consortium is ADS competency, which will inform and accelerate a corresponding federal motor vehicle safety standard," Administrator Jonathan Morrison said at the AI Horizons Summit in Pittsburgh [9].
NHTSA says the framework it is building will put safety at the foundation without stifling innovation [6], and Morrison said this week it should keep the agency from playing "whack-a-mole" with policy decisions [7]. A consortium is not a rule, and he was direct about where the rule comes from. "Ultimately, our ADS performance rulemaking will provide federal leadership in this space," he said [16]. He also tied the competency standard to the "burgeoning past patchwork of inconsistent state and local level requirements" [10]. For an operator running in several cities, one federal bar in place of several local ones would cut compliance work. The newsletter does not say what competency will measure.
The deals are not waiting. Lucid Motors and the European platform Bolt have agreed to "deploy at least 25,000 fully autonomous vehicles" on Lucid's coming midsize EV architecture [18]. May Mobility, which runs autonomous Toyota Siennas in three locations, plans to go public through a merger with ACP Holdings Acquisition Corp. in a deal that could raise more than $300 million at a $1.4 billion valuation [12][13]. The raise is roughly 21 percent of that valuation [15].
I'd expect the buyer, a city or a transit agency, to check the numbers it can verify itself and discount the rest.
So there are two questions worth running against every safety number a team publishes, before ASCEND picks its metrics. Whether an outside party could reproduce it from data the fleet already logs, and whether the team would still publish it if someone else chose the metric and the route. Figures that pass both are the ones an ADS competency standard can use, and the collection work is already paid for.
What to watch
- ASCEND's first work product: a published metric list tells fleet teams what to instrument, a charter tells them nothing.
- Whether NHTSA's ADS performance rulemaking includes language displacing state and local AV requirements, as Morrison's remarks suggest.
- Whether the May Mobility SPAC closes at the stated valuation, and what safety performance data the merger filings include.