2019/12/23 Strategy Analytics - NHTSA & The End of Expertise

Two years ago, at CES2018, Nvidia proudly announced its strategic relationship with Uber to deploy self-driving technology using Nvidia-supplied GPUs. Three months later a now-famous fatal crash in Phoenix involving an Uber self-driving Volvo led to a recently-concluded investigation by the National Transportation Safety Board and still-unfolding consequences for the autonomous vehicle industry.

The fatal crash in Phoenix occurred in the context of a minimal regulatory regime in Arizona designed to attract tech innovators and startups including the likes of both Uber (pursuing ride hailing and autonomous vehicle tech) and Waymo. Arizona, and other states including Florida and California, have been able to create relatively liberal regulatory guidelines in the absence of any Federal oversight or action.

The U.S. regulatory environment in the automotive industry has been almost completely neutered under the Trump administration which has made de-regulation a focal point – going to far as to require at least two de-regulatory steps for any new regulatory action. This has more or less terminated any efforts focused on regulating autonomous vehicles.

In fact, the de-regulation actions have only accelerated the brain drain at the National Highway Traffic Safety Administration, where the outflow of senior executives to tech companies like Google, Waymo, Uber, and others continues apace. The net effect is that an already embattled agency – just beginning to recover from the tsunami of investigations following Toyota unintended acceleration, General Motors ignition switches, Takata airbags, and Volkswagen diesel cheating - is now doing battle with the executive branch – and is led by an acting administrator in the third year of the aforementioned Trump administration.

The frustration with the agency is evident in the NTSB report which supplies some notable commentary on the Uber event including:   

"The automated vehicle (AV) event that resulted in a fatally injured pedestrian was not an accident. Accidents are unpredictable and unforeseeable. Rather, this was a crash that was predictable and avoidable." - Bruce Landsberg, Vice Chairman NTSB

On automation complacency: "Automation performs remarkably well most of the time and therein lies the problem. Human attention span is limited, and we are notoriously poor monitors.”

On driver inattention: "...the safety driver spent 34% of the time looking at her cell phone while streaming a TV show. The longest inattentive period was 26 seconds and in the 3 minutes prior to the crash, she glanced at the phone 23 times!...Automation complacency!"

On tech: "Uber ATG started its test program in Tempe, AZ, with 2 humans aboard... ATG decided that a dedicated observer was not needed. The vehicle was performing as well or better than expected. Automation complacency!"

On NHTSA’s inaction: "...we chastised the NHTSA for not providing definitive leadership to the states to manage the expansive growth in AV testing. It’s time for NHTSA to live up to its stated goals & create appropriate safety regulation in this developmental area."

Link to pdf: https://www.ntsb.gov/investigations/AccidentReports/Reports/HAR1903.pdf

There is a reason the U.S. transportation industry is supported by an agency called the Department of Transportation. The skilled professionals and safety investigators and experts at USDOT and sub-agencies such as the FAA, FRA, and the NHTSA are in place to help instill confidence in the nation’s transportation networks.

The deregulatory push that has relegated the NHTSA to the sidelines regarding automated driving, likely places the onus for oversight onto the legislative branch. In this instance, Congress has in fact considered legislation to guide autonomous vehicle development, but the bill failed in the Senate. A renewed push for AV legislation is underway, but observers give this effort a less than 50% chance of success given the level of conflict between legislators and the determined blocking efforts of lobbyists.

The fatal Phoenix crash occurred in the context of an already over-matched regulatory agency facing further erosion from an Administration hostile to regulatory activity of any kind. The NHTSA is facing perhaps its greatest challenge ever in guiding the onset and adoption of active safety systems, but it is doing so in an environment that is allergic to experts and expertise.

Regulations have the power to provide legal protection for both innovators and the general public. Both parties have an interest in agreeing to a framework for testing automated cars and trucks on public roads. In the current environment, there is likely to be no protection and that is likely to undermine enthusiasm for developing these technologies.

Thankfully for the general public and the autonomous vehicle industry itself, there have been no further fatalities resulting from the use of fully automated driving systems. Tesla Motors continues to be implicated in crashes of vehicles using its semi-autonomous technology – cynically branded Autopilot. 

For its part, Nvidia has since de-emphasized its efforts at supporting robotaxi development, shifting its focus toward what it calls “Level 2+” semi-automated systems. In a recent Webinar on Nvidia GTC China the company noted:

• Pre-trained inference models are to become “open source” from the NGC (NVIDIA GTC Cloud) – but not NVIDIA’s actual data sets.
• NVIDIA will also offer its Driver Federated Server in which AD models and data can be shared within each market region and ensures IP protection (e.g. in a Chinese joint venture) and data privacy (as previously learnt from the medical sector).
• The new Drive AGX Orin SoC offering x7 the performance of Xavier with 200 TOPS, 17 trillion transistors. X3 performance per Watt than Xavier, around 60-70 Watts per TOPS, promising less thermal and more compact ECUs. Sampling dates and memory specs to come later.
• DiDi is to become the latest user of NVIDIA AI for AD development.

Nvidia will have no keynote and no booth at CES2020. The company is mostly launching products on gaming. The 2020 GTC event in Silicon Valley will be the largest yet. Maybe DiDi will succeed where Uber failed. Maybe China will take the lead in autonomous vehicle development. Maybe. I wonder what the experts think.