Employees at the University of Central Florida conducted a study involving more than 37,000 vehicles to analyze accident statistics. It turns out that cars equipped with autopilot are less likely to have accidents, but there are exceptions.
Scientists turned to a statistical model called logistic regression. During the research work, parameters such as road conditions, weather, time and place of accidents were taken into account.
In most cases, it is safer to drive a car with an autopilot because the computer is able to recognize road objects and avoid them prematurely with more precise control.
But in some cases, self-driving cars are not as good as cars with drivers. For example, at dawn or dusk, that is, at a time when there is not enough light. This circumstance increases the likelihood of an accident involving an autopilot by up to 5 times.
Experts attribute this gap to the fact that modern intelligent systems are not programmed for complex driving scenarios in limited lighting conditions. To solve this problem, cars need to be equipped with more advanced sensors.