(Repeats MARCH 21 story. No change to text.)
By John Kemp
LONDON, March 22 (Reuters) - Uber has suspended road tests of self-driving vehicles after the first pedestrian was killed by one of its vehicles operating under autonomous computer control.
Proponents of self-driving vehicles claim they would be safer, as well as more energy efficient, but those safety claims will now come under heightened scrutiny.
In theory, self-driving vehicles should eliminate several major driver-related causes of road traffic accidents, including excessive speed, intoxication and inattention.
But there are concerns about how self-driving vehicles will interact with unpredictable human drivers and pedestrians as well as unmapped hazards such as temporary road obstructions.
There are also fears about how self-driving vehicles would cope with hacking or widespread disruption of their communications systems (“Too much sun could wreak havoc on driverless cars”, Bloomberg, March 16).
The accident investigation into Sunday’s crash will mark an important test of the technology but also of the ability of politicians, regulators and the media to think about risk in an intelligent way.
All transportation is fraught with risk and road transport is particularly dangerous even with a human behind the wheel (“Self-driving Uber cars kills Arizona woman crossing street”, Reuters, March 20).
Every year, there are more than 6 million reported crashes involving motor vehicles in the United States, according to the Bureau of Transportation Statistics (“Motor vehicle safety data”, BTS, 2017).
In 2016, almost 40,000 people were killed in transport-related accidents, according to the National Transportation Safety Board (“U.S. transportation fatalities 2015-2016”, NTSB, 2017).
Of those, more than 95 percent were killed in accidents involving motor vehicles (37,461), dwarfing the number killed on the railroads (733), marine transport (730), aviation (412) and pipelines (16).
Road traffic fatalities included nearly 6,000 pedestrians, over 5,000 motorcyclists and 840 pedal-cyclists (“Fatal motor vehicle crashes overview”, U.S. Department of Transportation, 2017).
Road traffic crashes are the number one cause of death for young people between the ages of 8 and 24, according to the National Highway Transportation Safety Administration.
And they are one of the top two causes of accidental death for people in all age groups (“Motor vehicle traffic crashes as a leading cause of death in the United States”, NHTSA, 2018).
In a worrying trend, U.S. roads appear to have become more dangerous in the last few years, reversing the downtrend over the previous quarter of a century.
The number of crashes, injuries and fatalities has been growing faster than highway traffic volumes since around 2015.
Road deaths rose to 1.18 per 100 million miles driven in 2016 up from 1.15 in 2015 and a recent low of 1.08 in 2014.
The trend among pedestrians has been particularly worrying, with fatalities surging by 9 percent in 2016 to the highest level since 1990.
Many road accidents have been attributed in whole or part to driver misbehaviour. Alcohol was involved in more than 10,000 fatalities in 2016. Speed in more than 10,000. And distraction in more than 3,000.
Driving is already relatively dangerous in the United States, so the question is whether self-driving vehicles increase or reduce the existing risks, not whether they can eliminate risk altogether.
So far, self-driving vehicles have completed fewer than 10 million miles of on-highway tests (“Self-driving cars under scrutiny after pedestrian death”, Financial Times, March 19).
Supporters and critics have long acknowledged the inevitability of a self-driving vehicle being involved in a serious accident, and eventually being the cause of a fatality.
Some had hoped self-driving vehicles would log 100 million miles or more before the first fatal accident to prove they are safer.
But that shows a misunderstanding of how statistics work, since the first fatal accident is as likely to occur on the first mile as on the 100-millionth mile (assuming the safety level remains constant).
Until self-driving vehicles are tested on actual roads for tens of millions or even hundreds of millions of miles it will not be possible to evaluate their safety relative to human-operated vehicles.
The advent of self-driving vehicles poses complicated questions about safety but is no different from any other new technology.
Every new technology is initially beset by accidents, and that has been particularly true in transportation, where early steam engines, automobiles and aircraft were all initially accident-prone.
All federal safety regulation in the United States stems originally from the response to the problem of exploding steam boilers in the nineteenth century (“Bursting boilers and the federal power”, Burke, 1966).
The advent of high-pressure steam engines on riverboats prompted a spate of explosions killing crew, passengers, and unlucky bystanders.
Between 1818 and 1824, 47 lives were lost in 15 boiler explosions on steamboats. Between 1825 and 1830 there were 42 explosions killing 273 people.
One particularly serious explosion aboard the steamboat “Helen McGregor” near Memphis reportedly killed 50-60 people.
“The many distressing accidents which have of late occurred in that portion of our navigation carried on by the use of steam power deserve the immediate and unremitting attention of the constituted authorities of the country,” President Andrew Jackson wrote in his State of the Union message to Congress in 1833.
“The fact that the number of those fatal disasters is constantly increasing, notwithstanding the great improvements which are everywhere made in the machinery employed and in the rapid advances which have made in that branch of science, shows very clearly that they are in a great degree the result of criminal negligence.”
“That these evils may be greatly lessened, if not substantially removed, by means of precautionary and penal legislation seems to be highly probable.”
The eventual response after many years of delay was a system of safety regulation including boiler standards and inspections.
The arrival of the internal combustion engine and the automobile generated a similar surge of accidents among drivers and pedestrians and a legal and regulatory response.
Early aircraft, too, were far more prone to mechanical failure or pilot error than has been the case in recent years.
But no one would now suggest steam boilers, automobiles or aircraft are simply too dangerous to be practical - any more than in future people will think self-driving vehicles are too dangerous.
In every case of new technology, safety lessons have been learned through a painful process of trial and error that has unfortunately cost lives.
Self-driving vehicles will be no different.
The challenge is to introduce sensible and balanced safety regulations while encouraging continued development of a technology that has potential to save lives and bring a range of other benefits.
That requires a very transparent approach from the technology companies and a calm and reasoned response from politicians, regulators and journalists. (Editing by Mark Potter)