Federal judge dismisses GM ignition switch case
Study finds self-driving cars repeatedly fail to stop for obstacles
Self-driving cars have been found to not stop for certain obstacles or have presented other dangerous situations unless a driver takes immediate manual control, according to a new study from the Insurance Institute for Highway Safety (IIHS).
During a sequence of on-road and track tests for BMW, Mercedes-Benz, Tesla and Volvo, IIHS researchers discovered that self-driving cars repeatedly either failed to recognize obstacles or failed to stop when presented with one.
Other times, the cars would crash in response to an obstacle if the driver didn't take over, IIHS said.
"The early results underscore the fact that today's systems aren't robust substitutes for human drivers," the insurance trade group said in its report published Tuesday.
In particular, researchers found that the BMW, Volvo and Mercedes self-driving cars did not brake when a vehicle stopped ahead of them in certain circumstances.
Tesla's models also failed to respond to obstacles under some conditions, hitting a balloon used as a target.
The research also found that drivers' expectations that they are safe in self-driving cars fuels the tendency for the automobiles to crash.
IIHS chief researcher David Zuby said that this presents a real problem for designers.
"If they limit functionality to keep drivers engaged, they risk a backlash that the systems are too rudimentary. If the systems seem too capable, then drivers may not give them the attention required to use them safely," he said.
The researchers expressed caution about the viability of testing self-driving vehicles on real roads, pointing to the incident last March when a self-driving Uber prototype hit and killed a pedestrian.
"The Uber crash in Arizona that took the life of a pedestrian in March shows the hazards of beta testing self-driving vehicles on public roads," the report said.