Autonomous Vehicle Engineering - September 2021 - 3

The Navigator
What will result from NHTSA's Tesla
Autopilot investigation?
More than five years after Joshua Brown's
death, life signs are emerging at the National
Highway Traffic Safety Administration (NHTSA).
Brown was decapitated when his Tesla Model S
drove itself under a tractor-trailer while he wasn't
paying attention. In the intervening years, there
have been dozens of crashes where Tesla's Autopilot
advanced driver-assistance system (ADAS) setup
was active. Until recently, NHTSA has done nothing
to reign in how Tesla and other automakers deploy
such systems. Just weeks after the agency issued
an order for all companies developing
or deploying partially or highly automated
driving systems to report all
crashes, it has opened (as of this
writing in mid-August) a formal
investigation into Teslas slamming
into parked emergency vehicles.
Sadly,
the meme of Tesla's
radar and one or more cameras (although Tesla
recently stopped using radar in North America).
The radar tracks the distance and closing speed to
the car ahead to maintain a safe gap for adaptive
cruise control (ACC). However, the narrow beam
and limited resolution make it difficult to distinguish
between objects.
As a result, at speeds above about 40 mph (64
plowing into the back of fire trucks
and police cars on the side of the
road has become something of a
running joke in the last few years.
The new NHTSA investigation is
specifically focused on at least 11 such
incidents. Chances are similar that incidents have
occurred with other OEMs' vehicles as well, but at
this point we simply don't have the data to know.
I recently asked General Motors' Duncan Aldred,
the global VP of GMC and Buick, if there have been
any crashes with vehicles where the automaker's
hands-free Super Cruise system was active. He
sidestepped the question, providing no response.
While the new inquiry by NHTSA's Office of
Defect Investigations is specifically targeting Tesla,
most automakers now are offering ADAS with
similar functionality to Autopilot. It's possible that
whatever result comes from this investigation may
impact the entire industry and how it deploys ADAS.
It may even result in new rulemaking by the agency.
Like most such systems, Autopilot has used a
combination of long-range/low-resolution forward
Ideally, NHTSA
will bar
automakers
from publicly
beta testing
their safetycritical
software
in the hands of
consumers.
km/h), any signal returns from objects that appear
to be stationary are typically ignored. In most cases,
when using ACC for its intended use on highways,
stationary objects usually are road
signs and overpasses. Responding
to all of these returns would yield
repeated phantom braking, another
common complaint from Tesla drivers.
The inconsistent performance of
Tesla's camera-based machine vision,
combined with its limited ability to
gauge distance and position of
objects, has led to numerous crashes
when emergency vehicles are
stopped but partially intruding into
the roadway. It's likely that systems
deployed by other automakers are
also susceptible - but there is a key difference.
Tesla drivers seem instilled with a much
higher and unwarranted degree of confidence in
the performance of Autopilot. They seem more
inclined to use it hands-free and eyes-off. Without
an IR-camera-based driver monitor and capacitive
sensors in the steering wheel to minimize abuse,
it's difficult to curtail inattentive driving.
Perhaps the NHTSA investigation will lead
to rulemaking that mandates robust driver monitoring
for all automakers. That also would have more
impact on reducing distracted or drowsy driving
than any laws banning use of handheld devices.
Ideally, NHTSA will also bar Tesla and other automakers
from publicly beta testing their safety-critical
software in the hands of consumers. It's a longshot,
but at least NHTSA now is doing something. ■
Sam Abuelsamid
Principal Analyst,
eMobility
Guidehouse Insights
sam.abuelsamid@
guidehouse.com
AUTONOMOUS VEHICLE ENGINEERING
September 2021 3
Autonomous Vehicle Engineering - September 2021

Table of Contents for the Digital Edition of Autonomous Vehicle Engineering - September 2021
