Professional Documents
Culture Documents
Reality
check
Research, deadly crashes show need
for caution on road to full autonomy
SPECIAL ISSUE:
AUTONOMOUS VEHICLES
Vol. 53, No. 4 | August 7, 2018
The road to full driving autonomy is paved with good intentions:
Reduce crashes, reduce deaths, reduce congestion, increase
mobility. That is the bright future the industry is chasing. Like
the evolution of any nascent technology, however, there have
been glitches and misfires along the way. Even deaths.
The idea of self-driving cars has gar- New IIHS research based on track tests The Uber crash in Arizona that took the
nered so much press that consumers can and on-road experiences with Level 2 life of a pedestrian in March shows the
almost be forgiven for thinking the latest driver assistance uncovers some of the in- hazards of beta testing self-driving vehi-
cars can drive themselves. herent challenges with partial automation. cles on public roads. IIHS researchers ex-
While it is true that many new vehi- The deadly crash of a Tesla Model X plore how automatic emergency braking
cles can assist drivers in performing cer- on a California highway in March demon- and better headlights might have helped
tain tasks, such as maintaining following strates the limits of the technology and the prevent this tragedy.
distance and lane centering, no car can propensity of some drivers to misuse it. Finally, a patchwork of state laws and
handle every driving task on a full range of A HLDI analysis of Tesla insurance voluntary federal policy guidelines lacks
roads and conditions. losses reveals benefits for the combined the safeguards needed to protect every-
This special issue of Status Report is a crash avoidance features on the Model S, one on the road as fully autonomous vehi-
follow-up to the November 2016 special while the benefit of adding “Autopilot” is cles are tested and eventually deployed in
issue on autonomous vehicles. limited to lowering collision claims. the U.S. n
August 7, 2018 |3
(« from p. 3) function better than they do “At IIHS we are coached to intervene cross the line on the inside of the curve in
in more complex driving situations. without warning, but other drivers might one trial. None of the other systems tested
A case in point is the stopped-vehicle not be as vigilant,” Jermakian says. “ACC provided enough steering input on their
ACC tests. On the track, the 5 series, E- systems require drivers to pay attention to own to consistently stay in their lane, often
Class and Teslas braked to avoid the target what the vehicle is doing at all times and be requiring the driver to provide additional
vehicle. This was the case even though the ready to brake manually.” steering to successfully navigate the curve.
owner’s manuals for all the test vehicles Unnecessary or overly cautious brak- The E-Class stayed within the lane in 9 of
warn that ACC may not brake when it en- ing is an issue IIHS noted in the Model 3. 17 runs and strayed to the lane marker in
counters vehicles that are already stopped In 180 miles, the car unexpectedly slowed five trials. The system disengaged itself in
when they come into sensor range. down 12 times, seven of which coincided one trial and crossed the line in two. The 5
Out on the road, engineers noted instances with tree shadows on the road. The others series stayed within the lane in 3 of 16 trials
in which each vehicle except the Model 3 were for oncoming vehicles in another lane and was more likely to disengage than steer
or vehicles crossing the road far ahead. outside the lane. The S90 stayed in the lane
“The braking events we observed didn’t in 9 of 17 runs and crossed the lane line in
create unsafe conditions because the decel- eight runs.
erations were mild and short enough that When trying out new vehicles in hilly
the vehicle didn’t slow too much. However, Central Virginia, home to the VRC, engi-
unnecessary braking could pose crash risks neers noted early on that advanced driver
in heavy traffic, especially if it’s more force- assistance systems that rely on seeing road
ful,” Jermakian says. markings to keep vehicles in their lanes
“Plus, drivers who feel that their car were sometimes flummoxed by hills. As a
brakes erratically may choose not to use vehicle crests a hill, the lane markers on the
adaptive cruise control and would miss out road beyond are obscured.
on any safety benefit from the system.” For the on-road tests, engineers mapped
The outlook is promising for the poten- out a course that included three hills with
tial safety benefits of ACC. The technolo- different slopes. Drivers made six trial runs
gy is often bundled with forward collision on each hill in each vehicle.
warning and autobrake, and research by The E-Class stayed in its lane in 15 of 18
IIHS and HLDI has found crash-reduction trials and on the line in one trial, contin-
benefits for these systems combined. A fed- uously providing steering support without
erally sponsored study found that driv- erratic moves when lane lines weren’t visi-
ers using ACC have longer, safer following ble. The Model 3 also stayed in the lane in
distances than drivers who don’t use ACC. all but one trial, when it hugged the line.
Still, IIHS tests indicate that current ACC In contrast, the 5-series, Model S and S90
systems aren’t ready to handle speed con- struggled. The 5-series steered toward or
trol in all traffic situations. across the lane line regularly, requiring driv-
ers to override the steering support to get
IIHS can’t say yet which company has the Active lane-keeping it back on track. Sometimes the car disen-
safest implementation of Level 2 driver Engineers focused on two situations that gaged steering assistance on its own. The car
assistance, but it is important to note challenge active lane-keeping systems — failed to stay in the lane on all 14 valid trials.
that none of these vehicles are capable of curves and hills — in tests on open roads The Model S was errant in the hill tests,
driving safely on their own. A production with no other vehicles around. They also ob- staying in the lane in 5 of 18 trials. When
autonomous vehicle that can go anywhere, served how the systems performed in traffic. cresting hills, the Model S swerved left and
anytime isn’t available at the local car All five systems provide steering assis- right until it determined the correct place
dealer and won’t be for quite some time. tance that centers the vehicle within clearly in the lane, jolting test drivers. It rarely
marked lanes. They also may use a lead ve- warned them to take over as it hunted for
failed to respond to stopped vehicles ahead. hicle as a guide when traveling at lower the lane center. The car regularly veered
Jermakian recounts her experience with speeds or when the lead vehicle is blocking into the adjacent lanes or onto the shoulder.
the E-Class on U.S. 33 near the IIHS-HLDI the system’s view of the lane markers ahead. When drivers intervened to avoid poten-
Vehicle Research Center (VRC). Traveling To test active lane-keeping on curves, en- tial trouble, the active lane-keeping system
about 55 mph with ACC and active lane- gineers conducted six trials with each vehi- disengaged. Steering assistance only re-
keeping engaged but not following a lead cle on three different sections of road with sumed after drivers re-engaged Autopilot.
vehicle, the E-Class system briefly detect- radii ranging from 1,300 to 2,000 feet. The S90 stayed in the lane in 9 of 16
ed a pickup truck stopped at a traffic light Only the Model 3 stayed within the lane trials. The car crossed the lane line in two
ahead but promptly lost sight of it and con- on all 18 trials. The Model S was similar but trials and in four trials disengaged steer-
tinued at speed until she hit the brakes. overcorrected on one curve, causing it to ing assistance when it crested hills but
August 7, 2018 |5
Photo by Florida Highway Patrol investigators
The first fatal crash in the U.S. of a Tesla in “Autopilot” mode occurred in Florida in May 2016.
Neither the Model S (above) nor the driver braked for a tractor-trailer crossing the car’s path.
Several other Tesla crashes have made headlines, including one in Laguna Beach, Calif. (right).
August 7, 2018 |7
(« from p. 7) and after Autopilot was en-
abled instead of comparing vehicles with
and without the system over a specific time
Fatal Uber crash shows risks
frame. The pre-Autopilot period includ-
ed only the nine months of data after Tesla
activated forward collision warning and
of testing on public roads
S
before it enabled Autopilot. elf-driving cars are supposed to the dash cam video. The operator told in-
In this limited analysis, HLDI found that be better at averting crashes than vestigators that she had been monitoring
the frequency of claims filed under PDL, human drivers, but tests of proto- the system interface in the center console.
BI, MedPay and PIP didn’t change once type vehicles on public roads so far indi- Tempe police in June released a report in-
Autopilot was enabled, but the frequency cate that they aren’t always up to the task. dicating that the operator’s smartphone
of collision claims fell by 13 percent. Absent regulatory oversight, the race to was streaming a TV show in the 42 min-
“To get a better picture of how Autopi- deploy autonomous vehicles risks jeopar- utes preceding the crash.
lot is affecting claims, we need more data dizing public trust and safety and the life- The NTSB report indicates that the Uber
on how many Teslas are equipped with Au- saving promise of the technology. self-driving system first detected Herz-
topilot and how often it is used,” says Matt The National Transportation Safety berg about 6 seconds before impact, ini-
Moore, HLDI’s senior vice president. “The Board (NTSB) on May 24 issued a four- tially classifying her as an unknown object,
reductions in the frequency of third-party page preliminary report on the first fatal then a vehicle and then as a bicycle. At 1.3
crash involving a pedestrian and a self- seconds before impact, the report states,
driving vehicle operating under the control the system “determined that an emergency
of a computer, not a person. braking maneuver was needed to mitigate a
Until the March 18 tragedy, Uber Tech- collision.” Less than a second before impact,
nologies Inc. had done extensive testing the test operator took the wheel and started
of its fleet of prototype autonomous vehi- to brake just after hitting Herzberg.
cles in Arizona. That testing has since been The Uber self-driving system was operat-
shelved in the state. ing normally at the time of the crash, with
The details of the March 18 crash in no faults or diagnostic messages.
Tempe, Arizona, are by now well-known. “What’s chilling is that the engineers
What wasn’t known publicly until the behind Uber’s software program disabled
NTSB report’s release, however, is that the system’s ability to avoid a life-or-death
while an Uber experimental vehicle op- scenario while testing on public roads,” says
erating in self-driving mode is capable of David Zuby, the Institute’s chief research
detecting impending conflicts, it isn’t pro- officer. “Uber decided to forgo a safety net
grammed to brake or warn the test opera- in its quest to teach an unproven computer-
tor to take action. control system how to drive.”
“According to Uber, emergency braking
maneuvers are not enabled while the ve- Autobrake with pedestrian detection
hicle is under computer control, to reduce Institute staff have logged more than
the potential for erratic vehicle behavior,” 80,000 combined highway miles behind
the NTSB report states. “The vehicle op- the wheels of cars and SUVs equipped with
erator is relied on to intervene and take advanced driver assistance technologies to
action. The system is not designed to alert gauge the performance and quirks of each
the operator.” system and how drivers interact with and
physical damage and injury liability claims Elaine Herzberg, 49, was walking her bi- view them (see Status Report, March 29,
associated with Tesla’s version 1 hardware cycle across a four-lane arterial road around 2018, and Nov. 10, 2016, at iihs.org).
are in line with the benefits HLDI has doc- 10 p.m. when she was struck by a 2017 Volvo For the past several years, IIHS and
umented for comparable systems from XC90 modified with Uber’s sensors and HLDI researchers have studied the crash
other manufacturers.” software to operate in autonomous mode. avoidance technologies that are the precur-
Moore adds, “When we evaluated Teslas Herzberg had crossed more than three lanes sors of autonomous driving systems, ana-
with the version 1 hardware after the Au- before she was struck by the SUV at about lyzing data in insurance claims and police
topilot software was deployed, we saw a 39 mph. Dash cam video released by Tempe reports and conducting test track, on-road
significant reduction in collision claim fre- police shows Herzberg didn’t look in the and lab evaluations (see Status Report, Feb.
quency but no other changes.” SUV’s direction until just before it hit her. 22, 2018, Aug. 23, 2017, June 22, 2017, Nov.
For a copy of HLDI Bulletin Vol. 34 No. In the Uber, the lone test operator at the 17, 2016, and Nov. 10, 2016).
30, “Tesla Model S driver assistance tech- wheel wasn’t watching the road just before The XC90 is among the vehicles IIHS re-
nologies,” email publications@iihs.org. n impact, according to the NTSB report and searchers have tested.
The ability for drivers to see the road ahead at be in almost total darkness. The pedestrian
night — and other drivers and pedestrians to was wearing dark clothes, and the bicycle she
see oncoming vehicles, too — is an important was pushing didn’t have side reflectors. There
area of IIHS research that may have come into are streetlights along this section of road, but
play in Tempe. the crash site wasn’t directly illuminated.
About half of traffic deaths occur either in According to the NTSB preliminary report,
the dark or at dawn or dusk, and the propor- Uber’s lidar and radar first detected Herzberg extra time to see the pedestrian and act to
tion of pedestrians killed in low light condi- 6 seconds before impact but didn’t know what avoid the crash or lessen its severity.
tions is even greater. It is crucial that drivers, to make of her. It is possible that with better Crash reports don’t indicate whether the
whether human or machine, have a good view lighting the cameras could have helped con- XC90’s low beams or high beams were in use.
of the road at night to drive safely. That is the firm she was a pedestrian. The SUV has high-beam assist, which auto-
role of headlights, especially on roads without Velodyne Lidar Inc., which supplies the sen- matically switches between high beams and
street lighting. sors Uber uses, says lidar was capable of low beams, depending on the presence of
The Uber that struck and killed Elaine Her- detecting a pedestrian with a bicycle, but deci- other vehicles. Research shows drivers rarely
zberg had a variety of sensors to help it “see” sions about whether to brake or take evasive turn on their high beams. High-beam assist
the road and its surroundings. These includ- action were left to Uber’s software. ensures that they do.
ed light detection and ranging (lidar) sensors, The crash involved a specially outfitted 2017 “Headlights probably don’t come to mind
radar sensors and cameras. While lidar and Volvo XC90. Its headlights are rated poor be- when you think of autonomous vehicles, but
radar sensors don’t depend on ambient light cause they don’t provide sufficient low-beam they are important safety equipment, and we
to see, cameras, like human eyes, do. light in IIHS evaluations. Good-rated headlights intend to continue our evaluations to encour-
The high-contrast video recorded by the would have illuminated twice as much of the age automakers to improve them,” IIHS Presi-
Uber dash camera makes the road appear to road ahead for an attentive driver. That means dent David Harkey says. n
August 7, 2018 |9
(« from p. 9) given the system more time to
identify Herzberg as a pedestrian and Her-
zberg more time to finish crossing the road.
Lax U.S. oversight of industry
“At 6 seconds out, the automated driv-
ing system had a range of choices it could
make,” Zuby says. “Just like humans might
jeopardizes public safety
A
ignore something unexpected and unclear patchwork of state laws and volun- At the same time, an industry coalition
moving toward their path, Uber’s system tary federal guidelines is attempt- of automakers, suppliers, tech firms and
didn’t react despite sophisticated sensors ing to cover the testing and eventual other groups is urging swift passage of the
and artificial intelligence. That’s unaccept- deployment of autonomous vehicles in the bill in the name of safety and mobility.
able. To be better than human drivers, auto- U.S. It is a decidedly pro-technology ap-
mated systems have to make safer choices.” proach that lacks adequate safeguards to States forge ahead
protect other road users. As of August, 31 states and the District of
According to Uber, emergency braking ma- “We don’t want to hamstring the devel- Columbia have enacted legislation or taken
neuvers are not enabled while the vehicle opment of autonomous vehicles but do executive action on driving automation. The
is under computer control, to reduce the want to ensure that all motorists, bicyclists laws in 11 of those states authorize a study,
potential for erratic vehicle behavior, the and pedestrians sharing the road are pro- define key terms or authorize funding.
NTSB report states. The vehicle operator is tected,” says David Harkey, IIHS president. Nine states authorize testing, while 11
relied on to intervene and take action. “The industry needs to take precau- states and D.C. authorize full deployment.
tions when operating experimental vehi- Some states impose substantial restric-
Deadly inattention cles on public roads and should share data tions on operators developing, testing and
In lieu of intervention from an autobrake on crashes, near-crashes and system disen- deploying driverless cars on public roads,
system with pedestrian detection, an alert gagements with the public,” Harkey says. while others welcome testing and deploy-
human driver may have been able to slow Beyond issuing policy guidelines (see ment with little legislative or regulatory
down the vehicle enough to reduce the se- Status Report Nov. 10, 2016, at iihs.org), the burden/impediment.
verity of the crash. National Highway Traffic Safety Adminis- For example, four states require the oper-
In the dash cam video, the pedestri- tration (NHTSA) hasn’t attempted to regu- ator to carry at least $5 million in insurance
an appears about 80 feet before impact. A late self-driving vehicles. At the time of this or surety bond while testing or deploying
best-case driver reaction time is about one writing, legislation that would set parame- automated vehicles on public roads. Six
second. At 40 mph the XC90 would have ters on their development, testing and de- states and D.C. authorize testing or deploy-
traveled 60 feet before the driver initiated ployment was stalled in Congress. ment without imposing any special finan-
braking at 20 feet, reducing the SUV’s speed The U.S. House of Representatives passed cial requirements.
by roughly 10 mph and the impact speed to the SELF DRIVE Act in September 2017, California, Connecticut, Massachusetts,
30 mph. The driver still would have struck but the Senate version, dubbed the AV Michigan, New York, Ohio, Pennsylvania
the pedestrian, but the reduced speed may START Act, is on hold. The act would pre- and Washington require operators to share
have raised her chances of survival. empt state laws on autonomous vehicles and a safety plan or assessment and/or some
The video shows the Uber operator look- drastically lift caps on the number of vehi- level of data, such as disengagement reports
ing down approximately 9½ seconds of the cles sold each year that can be exempt from or incident records, with the state regulator.
12½-second video. At 40 mph, the SUV federal safety standards. California, Nevada and Pennsylvania also
covered 750 feet during that short period. A broad coalition of consumer and safety require a testing permit, while New York re-
The driver was distracted for all of the last advocates has appealed to Senate leaders quires testing under state police supervision.
5½ seconds, or 330 feet, before impact. not to advance the bill until the National Laws allowing the operation of auto-
Experimental studies have shown that Transportation Safety Board completes its mated vehicles initially required a human
drivers can lose track of what automat- investigations of recent fatal crashes. They operator to be present and capable of
ed systems are doing, fail to notice when say the bill lacks comprehensive safeguards, taking over in an emergency. However, 12
something goes wrong and have trouble sufficient government oversight and indus- states — Arizona, California and Mich-
retaking control. Some companies require try accountability. igan among them — allow testing or de-
two operators in autonomous vehicles un- The coalition calls for the bill to limit the ployment without a human operator in
dergoing testing on public roads. Uber had size and scope of exemptions from federal the vehicle, although some limit it to cer-
used two operators in earlier road tests but safety standards, provide for adequate data tain defined conditions. Nine states don’t
had recently scaled back to one. collection and consumer information, apply always require an operator to be licensed.
“Looking at the center stack takes the op- safety critical provisions to Level 2 systems,
erator’s eyes off the road,” Zuby points out. boost funding for NHTSA, ensure access Crash data should be public
“Two operators in the vehicle might have al- and safety for people with disabilities, and IIHS has encouraged regulators to re-
lowed one to monitor the system computer maintain state and local regulations absent quire companies to share information
and the other to focus on the road ahead.” n federal rules on automated vehicles. about every crash and disengagement of
August 7, 2018 | 11
Insurance Institute for Highway Safety
Highway Loss Data Institute
Status Report
SPECIAL ISSUE: AUTONOMOUS VEHICLES IIHS is an independent, nonprofit scientific and educational organization dedicated to reducing the losses — deaths, injuries and
property damage — from motor vehicle crashes.
IIHS tests will shape ratings program
HLDI shares and supports this mission through scientific studies of insurance data representing the human and economic losses
for advanced driver assistance features 42 resulting from the ownership and operation of different types of vehicles and by publishing insurance loss results by vehicle make
and model.
Tesla crash highlights risks
of partial automation45 Both organizations are wholly supported by the following auto insurers and funding associations:
HLDI report finds fewer claims for Model S MEMBER GROUPS The Main Street America Group
under certain coverage types 47 AAA Carolinas Mercury Insurance Group
Acceptance Insurance MetLife Auto & Home
Fatal Uber crash shows dangers of testing Alfa Alliance Insurance Corporation Mississippi Farm Bureau Casualty Insurance Company
self-driving vehicles on public roads 48 Alfa Insurance MMG Insurance
Allstate Insurance Group Munich Reinsurance America, Inc.
Better headlights may have helped American Agricultural Insurance Company Mutual Benefit Group
an attentive driver spot pedestrian 49 American Family Insurance Mutual of Enumclaw Insurance Company
American National Nationwide
Lax oversight of self-driving industry Ameriprise Auto & Home New Jersey Manufacturers Insurance Group
Amica Mutual Insurance Company NYCM Insurance
risks safety of other road users 410 Auto Club Enterprises Nodak Insurance Company
Auto Club Group Norfolk & Dedham Group
Vol. 53, No. 4 Auto-Owners Insurance North Carolina Farm Bureau Mutual Insurance Company
August 7, 2018 Bitco Insurance Companies Northern Neck Insurance Company
California Casualty Group Ohio Mutual Insurance Group
Celina Insurance Group Oregon Mutual Insurance Company
Censtat Casualty Company Paramount Insurance Company
CHUBB Pekin Insurance
Inquiries/print subscriptions:
Colorado Farm Bureau Mutual Insurance Company PEMCO Mutual Insurance Company
StatusReport@iihs.org Concord Group Insurance Companies Plymouth Rock Assurance
COUNTRY Financial Progressive Insurance
CSAA Insurance Group PURE Insurance
Copy may be republished with attribution. Desjardins General Insurance Group Qualitas Insurance Company
Images require permission to use. ECM Insurance Group Redpoint County Mutual Insurance Company
Elephant Insurance Company The Responsive Auto Insurance Company
EMC Insurance Companies Rider Insurance
Editor: Kim Stewart Erie Insurance Group Rockingham Insurance
Esurance RSA Canada
Writer: Sarah Karush Farm Bureau Financial Services Safe Auto Insurance Company
Art Director: Steve Ewens Farm Bureau Insurance of Michigan Safeco Insurance
Farm Bureau Mutual Insurance Company of Idaho Samsung Fire & Marine Insurance Company
Photographers: Steve Ewens,
Farmers Insurance Group SECURA Insurance
Craig Garrett, Dan Purdy, Kim Stewart Farmers Mutual of Nebraska Selective Insurance Company of America
Florida Farm Bureau Insurance Companies Sentry Insurance
Frankenmuth Insurance Shelter Insurance Companies
Gainsco Insurance Sompo America
/iihs.org GEICO Corporation South Carolina Farm Bureau Mutual Insurance Company
The General Insurance Southern Farm Bureau Casualty Insurance Company
Georgia Farm Bureau Mutual Insurance Company State Farm Insurance Companies
@IIHS_autosafety Goodville Mutual Casualty Company Stillwater Insurance Group
Grange Insurance Tennessee Farmers Mutual Insurance Company
Grinnell Mutual Texas Farm Bureau Insurance Companies
@iihs_autosafety Hallmark Financial Services, Inc. The Travelers Companies
The Hanover Insurance Group United Educators
The Hartford USAA
IIHS
Haulers Insurance Company, Inc. Utica National Insurance Group
Horace Mann Insurance Companies Virginia Farm Bureau Mutual Insurance
iihs.org/rss Imperial Fire & Casualty Insurance Company West Bend Mutual Insurance Company
Indiana Farm Bureau Insurance Western National Insurance Group
Indiana Farmers Insurance Westfield Insurance
iihs.org Infinity Property & Casualty
Kemper Corporation FUNDING ASSOCIATIONS
Kentucky Farm Bureau Mutual Insurance Companies American Insurance Association
Liberty Mutual Insurance Company National Association of Mutual Insurance Companies
Louisiana Farm Bureau Mutual Insurance Company Property Casualty Insurers Association of America