According to a study based on data from more than 2,000 self-driving vehicles, such cars may be safer than human-driven ones in some conditions, while potentially more dangerous in other conditions. The researchers said that in their work, compared to previous studies, they used more data from both autonomous and human-driven vehicles.
It was published in the journal Nature Communications study aimed to explore the differences in accident risk between autonomous vehicles and human-driven vehicles to inform future development of autonomous technology and improve safety. The study collected data on accidents involving 2,100 autonomous vehicles and 35,133 human-driven vehicles between 2016 and 2022.
The study, led by Shengxuan Ding, a doctoral student at the University of Central Florida, found that autonomous vehicles are likely to improve road safety, but there are still areas for improvement. According to the results, autonomous vehicles are less likely than humans to be involved in road incidents where the cause of an accident is an everyday situation, such as not keeping up with the flow of traffic or drifting out of the lane. It found that self-driving cars are more effective in avoiding rear and side collisions, which are 0.5 times and 0.2 times less likely, respectively.
According to the research, self-driving vehicles were on average 5.25 times more likely to crash in low-light conditions – early in the morning or at dusk, when the light is fading. They were also 1.98 times more likely to make a turning error than humans. The poor performance of autonomous vehicles in low-light conditions is attributed to “lack of situational awareness expected in complex driving scenarios and limited driving experience.” Improving the performance of autonomous vehicles in these conditions may require “advanced sensors, robust algorithms, and intelligent design considerations.”
Crashes were grouped by location and position of impact – most were rear-end at an intersection
According to the study, the sensors and cameras used by autonomous vehicles cannot quickly adapt to changes in lighting conditions, which can affect their ability to detect obstacles, pedestrians and other vehicles. At dawn and dusk, for example, shadows and reflections from the sun can confuse sensors, making it difficult to distinguish objects and identify potential threats. Furthermore, fluctuating light conditions can affect the accuracy of object measurement and recognition algorithms used by vehicles.
“Key factors include the development of weather and light sensors, the implementation of redundant measures and the efficient integration of sensor data. By focusing on these aspects, the safety of autonomous driving systems can be significantly improved in challenging situations,” the research states.
It should be noted here that the study covers data up to 2022. In the past two years, autonomous driverless taxis operated by Google-owned Waymo have been transporting more and more people and operating around the clock. They’re not perfect and they’re limited to a few cities, but there’s sure to be at least some progress in the last 24 months or so, and investors have been betting big on them.
Source: sg.hu
