The road accidents cost about 3% of START worldwide and are the leading cause of death in children and young adults. But since they are rare, it is difficult to identify the precise places most at risk. We rely most often on the number ofaccidents counted here. This leaves a lot of holes in the racket: an area is only identified at high risk if an accident has already occurred nearby.
But, thanks to theartificial intelligence, it may soon be possible to predict accidents before they even happen. Researchers from Massachusetts Institute of Technology (MIT) have built an algorithm of deep learning based on a dataset (traffic density, speed average, number of lanes, presence of pedestrians, road improvements, etc.) from records GPS and satellites.
The researchers then assessed relevance using statistics from 2017 and 2018, and verified that the software was well versed in predicting accidents. ” Our model is applicable to any city even in the absence of historical accident data », Argues Amin Sadeghi, researcher at Qatar Computing Research Institute (QCRI) and main author of study. This tool could naturally allow town planners to recommend road improvements that reduce risk, but also be integrated into applications like Waze Where Apple Maps to advise routes to avoid.
.
The road accidents cost about 3% of START worldwide and are the leading cause of death in children and young adults. But since they are rare, it is difficult to identify the precise places most at risk. We rely most often on the number ofaccidents counted here. This leaves a lot of holes in the racket: an area is only identified at high risk if an accident has already occurred nearby.
But, thanks to theartificial intelligence, it may soon be possible to predict accidents before they even happen. Researchers from Massachusetts Institute of Technology (MIT) have built an algorithm of deep learning based on a dataset (traffic density, speed average, number of lanes, presence of pedestrians, road improvements, etc.) from records GPS and satellites.
The researchers then assessed relevance using statistics from 2017 and 2018, and verified that the software was well versed in predicting accidents. ” Our model is applicable to any city even in the absence of historical accident data », Argues Amin Sadeghi, researcher at Qatar Computing Research Institute (QCRI) and main author of study. This tool could naturally allow town planners to recommend road improvements that reduce risk, but also be integrated into applications like Waze Where Apple Maps to advise routes to avoid.
.
