According to a recent study, self-driving cars cause fewer accidents than human drivers. But: The driverless vehicles are apparently bad at turning. The background.
Humans or autonomous cars: Who actually causes more accidents? Some people may assume that the accident rate for self-driving cars must be higher than for humans. However, a recent study suggests that driverless vehicles are safer under normal circumstances.
Self-driving cars drive more safely than humans
The study is one of the most comprehensive accident studies to date. It was carried out by scientists at the University of Central Florida. Shengxuan Ding and his colleague Mohamed Abdel-Aty collected accident data from 2,100 accidents in California and from the National Highway Traffic Safety Administration (NHTSA).
All cases involved vehicles equipped with self-driving or driver assistance technologies in some form. The researchers also used data on 35,000 accidents involving human drivers without assistance systems.
Using statistical matching, the research team looked for pairs of accidents in which incidents occurred under similar circumstances. They compared factors such as road conditions, weather and time of day. They also examined whether the accidents occurred at an intersection or on a straight road. They found 548 matches.
Self-driving cars can't turn so well
The analysis shows that vehicles with advanced driving systems are less likely to be involved in an accident in similar accident scenarios than vehicles driven by humans.
However, autonomous cars also seem to have some weaknesses. The analysis shows that self-driving vehicles have a five times higher accident risk than humans at dusk and in the dark. In addition, the accident rate when turning is almost twice as high.
“It is important to improve the safety of autonomous vehicles at dusk and when turning,” Shengxuan Ding told the science magazine New Scientist. “Key strategies include improving weather and light sensors and effectively integrating sensor data.”
Experts criticize missing data
How meaningful the results of such studies are is controversial in expert circles. Abdel-Aty also sees some obstacles. The database for accidents involving self-driving vehicles is still too small and limited.
Missy Cummings of George Mason University in Virginia shares this opinion. She said the number of accidents involving autonomous vehicles is so small that no blanket conclusions can be drawn about the safety performance of the technology.
In addition, many accidents are not reported to the police if they only involve minor damage to the bodywork. According to Eric Teoh of the Insurance Institute for Highway Safety in Virginia, this factor must also be taken into account. He conducted a study on Google's early tests with self-driving cars in 2017. The result: only three out of ten accidents appeared in police reports.
The study comes at a time when autonomous vehicles are increasingly taking over the streets, especially in various cities in the USA. For example, Google's spin-off Waymo is expanding its robotaxis operations in Austin, Los Angeles and San Francisco.
Also interesting:
Source: https://www.basicthinking.de/blog/2024/07/06/studie-selbstfahrende-autos/
