Autonomous Vehicle Engineering - September 2021 - 2

Editorial
Fool Self-Driving II
Certain to be the year's most significant
development in the automated-driving
milieu was the August announcement
from the National Highway Traffic Safety
Admin. (NHTSA) that it opened an investigation
into Tesla's Autopilot driver-assistance
system after a disturbing pattern
of accidents over the past several years.
To now, the agency's inaction on these
generally widely reported incidents has
mystified all manner of entities related to
automated-driving development.
On the facing page, our advanced-mobility
columnist Sam Abuelsamid
offers a succinct technical hypothesis
of a sensor-software mashup likely at
the root of the Tesla accidents. The
incidents under NHTSA investigation
- 11 since 2018 involving stationary
first-responder vehicles somehow
intruding in the Teslas' intended path
of travel - would seem to represent the
classic software " edge case " that even
laypeople can comprehend, no?
I posed that question to Michael
DeKort, an outspoken critic of current
industry thinking regarding development
of automated-driving systems
and one of the most widely educated
and tuned-in people I know regarding
everything automated-vehicle (AV),
from sensor design to simulation best
practices (a story from our September
2020 issue details DeKort's ideas
about AV simulation testing). As often
happens when we talk, his answer
offered stunning clarity.
" I don't think it is an edge case at
all. Edge cases should be statistically
rare events for a given ODD (operational
design domain). Stationary objects are
not rare. I would argue they are basic
scenarios. I think people confuse crashes
or unwanted scenarios or outcomes
with 'edge cases.' The outcome - or
whether a scenario or result is desired
- is irrelevant. It should be purely based
on statistical chance of occurrence. "
He added that there almost surely
would be more crashes in the Tesla situation
and other AV public-road trials,
but an attentive human is disengaging
and " saving " the system. Regarding
the 11 Tesla events under NHTSA investigation,
" It's actually far more, " DeKort
asserted. " The vast majority are hidden
by [system] disengagements. "
Mr. DeKort, who just joined SAE
International's task force considering
new revisions to the J3018 Standard
(Safety-Relevant Guidance for On-Road
Testing of SAE Level 3, 4, and 5 Prototype
Automated Driving System (ADS)Operated
Vehicles), may not be the
last word on Tesla's software, but in my
experience his observations and analyses
have proved accurate and prescient. If
NHTSA's experts are worth their salt, it's
probable the agency might reach conclusions
similar to his.
The implications for AV development
may be far-reaching, as nearly
all developers name safety as the
chief aspirational aspect of high-level
vehicle automation. At the automaker
level, many less " projective " than Tesla
about current system capabilities nonetheless
have been reticent to directly
correlate driver-assistance systems with
enhanced safety - largely because that
aspect, for now, is nigh on impossible
to prove. So, driver-assistance systems
thus far are billed mostly for their potential
to reduce driving " fatigue " or stress.
Whatever the result of NHTSA's
investigation of Tesla, one fundamental
tension between AV development and
AV marketing is likely to remain: Has the
customer been caused to expect too
much - or do current-technology driver-assistance
systems deliver too little?
Bill Visnic, Editorial Director
EDITORIAL
Bill Visnic
Editorial Director
Bill.Visnic@sae.org
Lindsay Brooke
Editor-in-Chief
Lindsay.Brooke@sae.org
Paul Seredynski
Senior Editor
Paul.Seredynski@sae.org
Ryan Gehm
Associate Editor
Ryan.Gehm@sae.org
Jennifer Shuttleworth
Associate Editor
Jennifer.Shuttleworth@sae.org
Lisa Arrigo
Custom Electronic
Products Editor
Lisa.Arrigo@sae.org
Contributors
Kami Buchholz
Detroit Editor
John Kendall
Europe Editor
Terry Costlow
Electronic Technologies Editor
Bradley Berman
U.S. West Coast Editor
Sebastian Blanco, Don Sherman,
Paul Weissler
DESIGN
Lois Erlacher
Creative Director
Ray Carlson
Associate Art Director
SALES &
MARKETING
Joe Pramberger
Publisher
joe@techbriefs.com
Kaitlyn Sommer
Marketing Director
ksommer@techbriefs.com
Martha Tress
Recruitment Sales Manager
+1.724.772.7155
Martha.Tress@sae.org
REGIONAL SALES
North America
New England/Eastern Canada:
ME, VT, NH, MA, RI, QC
Ed Marecki
+1.401.351.0274
emarecki@techbriefs.com
CT:
Stan Greenfield
+1.203.938.2418
greenco@optonline.net
Mid-Atlantic/Southeast/TX:
DC, VA, WV, TN, NC, SC, GA, FL,
AL, MS, LA, AR, OK, TX
Ray Tompkins
+1.281.313.1004
rtompkins@techbriefs.com
NY, NJ, OH:
Ryan Beckman
+1.973.409.4687
rbeckman@techbriefs.com
PA/DE:
Desiree Stygar
+1.908.300.2539
dstygar@techbriefs.com
Midwest/Great Lakes:
IN, MI, WI, IA, IL, MN
Chris Kennedy
+1.847.498.4520, x3008
ckennedy@techbriefs.com
Midwest/Central Canada:
KS, KY, MO, NE, ND, SD, ON, MB
Bob Casey
+1.847.223.5225
bobc@techbriefs.com
Southern CA, AZ, NM,
Rocky Mountain States:
Tim Powers
+1.424.247.9207
tpowers@techbriefs.com
Northern CA, WA, OR,
Western Canada:
Twyla Sulesky
+1.408.779.0005
tsulesky@techbriefs.com
International
Europe:
Sven Anacker
Britta Steinberg
+49.202.27169.11
sa@intermediapartners.de
steinberg@intermediapartners.de
China:
Alan Ao
+86.21.6140.8920
alan.ao@sae.org
Japan:
Shigenori Nagatomo
+81.3.3661.6138
Nagatomo-pbi@gol.com
South Korea:
Eun-Tae Kim
+82-2-564-3971/2
ksae1@ksae.org
Integrated Media
Consultants
Christian DeLalla
+1.973.841.6035
christiand@techbriefs.com
Patrick Harvey
+1.973.409.4686
pharvey@techbriefs.com
Todd Holtz
+1.973.545.2566
tholtz@techbriefs.com
Rick Rosenberg
+1.973.545.2565
rrosenberg@techbriefs.com
Scott Williams
+1.973.545.2464
swilliams@techbriefs.com
SUBSCRIPTIONS
+1.866.354.1125
AVE@OMEDA.COM
REPRINTS
Jill Kaletha
+1.574.347.4211
jkaletha@mossbergco.com
2 September 2021
AUTONOMOUS VEHICLE ENGINEERING
Autonomous Vehicle Engineering - September 2021

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