Survey: Self-driving cars don’t make roads completely safe

A new study suggests that while self-driving car technology is promising to reduce crashes, it may not be able to prevent all accidents caused by human error,media reported. Auto safety experts point out that 94 percent of all crashes in the U.S. are caused by human factors, but the American Highway Safety Insurance Association (IIHS) points out that self-driving cars controlled by computers can only stop a third of them.

Survey: Self-driving cars don't make roads completely safe


While self-driving cars will eventually identify danger faster than humans, react faster, and avoid distracted or drunk driving, it will be much more difficult to stop other crashes, the group said.

IIHS looked at more than 5,000 incidents collected by the National Highway Traffic Safety Administration, including those caused by misperceptions such as driver distraction, reduced visibility, or failure to detect hazards in a timely manner. The researchers also distinguished between “incapacity” caused by human drivers, such as drivers affected by alcohol or drugs, dozing drivers and drivers with health problems. Self-driving cars can prevent these conditions, the study found.

However, robotic cars may not be able to prevent other errors, including prediction errors, planning errors, execution errors, and so on.

Jessica Cicchino, co-author of the study, cites an example of a bicycle or another car that suddenly turns to a self-driving car, it may not stop quickly enough or turn in time. “Self-driving cars need not only to perfectly perceive the world around them, but also to react to what’s around them. “

Cicchino says how much of a car accident can be prevented depends largely on how self-driving cars are programmed. If self-driving cars comply with all traffic regulations, including speed limits, more traffic accidents will stop. She also says that if artificial intelligence makes them drive and react more like humans, it will reduce crashes.

“Building self-driving cars that can be driven like a human is a huge challenge in itself,” Alexandra Mueller, a research scientist at IIHS, said in a statement. But in reality they need to do better to deliver on the promises we’ve all heard. “

Partners for Automated Vehicle Education is an organization with many members of self-driving car companies. In response to the new study, the group said Thursday that it mistakenly believes that the only way self-driving cars can drive better than humans is through superior perception and distraction.

The group said research data showed that self-driving cars could prevent 72 percent of accidents or crashes, but the impact of these cars was too complex and the ultimate impact was only a guess.

Missy Cummings, a professor of robotics and human factors at Duke University who is familiar with the study, says that even if one-third of people are prevented from doing so, technology is too important. She points out that even vehicles equipped with laser, radar and camera sensors are unlikely to perform perfectly in all cases.

She points out that researchers and people in the self-driving car industry never thought the technology would prevent all of the crashes currently caused by humans. She said it was “a traditional outsider’s view that this technology will become a panacea that will prevent all deaths.” “

Cicchino said IIHS researchers reviewed the cause of the accident and decided what could be avoided, assuming that all vehicles on the road were autonomous. She said crashes when self-driving cars are mixed with human-driven cars will be prevented.

Survey: Self-driving cars don't make roads completely safe