Driverless cars ‘still prone to errors’
LONDON — Transport researchers have said that driverless cars may never be as safe as the best human drivers, and that computers will not be able to avoid all fatal accidents.
Michael Sivak and Brandon Schoettle from the University of Michigan Transportation Research Institute published a white paper on the safety of autonomous cars, which are currently being developed by academic institutions, Google and almost every mainstream car manufacturer.
Ironically, many of the accidents which computers will not be able to prevent are those caused by humans — either as pedestrians or drivers of traditional cars still using the roads.
Driverless cars could, however, reduce many of the 1,2 million road fatalities which the World Health Organisation estimates occur each year.
The paper acknowledges that driverless cars could help the elderly and disabled to remain independent and mobile, and also slash emissions by reducing congestion and idling. It also says that the safety performance of self-driving vehicles could, in principle, be made perfect. But it also points out that not all fatalities are caused by drivers — therefore, replacing them with a computer, even an infallible one, may not prevent all deaths.
The researchers give the example of a drunk man stepping out in front of a car with a “very short” gap.
If the limiting factor in that instance is not the reaction time of the driver, but the ability of the brakes to stop a car, then a driverless car “might not be able to stop in time”.
Another set of challenges involving other traffic participants requires recognizing and negotiating unusual road users.
Examples include ridden horses and horse drawn buggies, large non-automotive farm equipment, and situations where police or construction crews are required to direct traffic. Research has shown that around one per cent of fatal crashes in the US in 2013 involved some kind of component failure on the car as a contributing factor, another type of accident which driverless cars may not be able to prevent.
“Indeed, given the complexity of the sensing hardware and of the information-processing software, it is reasonable to expect that, overall, vehicular factors would likely occur more frequently on self-driving vehicles than on conventional vehicles,” the said.
The paper also warns of the period after the introduction of driverless cars, but before all manual cars had been replaced (the average age of a car in the US is 11.4 years) when interaction between the two could prove problematic.
“In many current situations, interacting drivers of conventional vehicles make eye contact and proceed according to the feedback received from other drivers. Such feedback would be absent in interactions with self-driving vehicles,” they said.
“The expectation of zero fatalities with self-driving vehicles is not realistic. It is not a foregone conclusion that a self-driving vehicle would ever perform more safely than an experienced, middle-aged driver. During the transition period when conventional and self-driving vehicles would share the road, safety might actually worsen, at least for the conventional vehicles.”
The University of Michigan is developing a 32-acre “mini-city” designed to test driverless cars, called M City. It includes junctions, roundabouts, pavements, bus stops and benches as well as simulated pedestrians and parked cars.
“Connected and automated vehicle technology will usher in a revolution in the mobility of people and goods comparable to that sparked by the introduction of the automobile a century ago,” said Peter Sweatman, director of the university’s Mobility Transformation Center.
“M City will allow us to rigorously test new approaches in a safe, controlled and realistic environment before we implement them on actual streets.” — Telegraph
Researchers from a Us university warn that driverless cars will never eradicate all fatal crashes