Data-driven approaches for road safety: A comprehensive systematic literature review

Road crashes cost over a million lives each year. Consequently, researchers and transport engineers continue their efforts to improve road safety and minimize road crashes. With the increasing availability of various sensor technologies to capture road safety-related data and the recent breakthrough in modern data-driven techniques, in particular Machine Learning and Deep Learning techniques, data-driven road safety research has gained significant attention in the past few years. As road safety involves a number of different aspects, including road infrastructure (e.g., surface conditions), road user behaviors (e.g., driver/pedestrian behavior), and traffic congestion, critically reviewing all these major aspects and their relationships with road crashes is a challenging task. In this paper, we present a detailed review of 70 articles, which are shortlisted from 2871 articles found by searching relevant keywords from the scopus IEEE digital library and google scholar databases. To better analyze the data-driven road safety research a number of taxonomies are first introduced to characterize data sources Equipment & sensors to capture data And methodologies to analyze and make decisions based on data. Then Based on the defined taxonomies Selected research articles covering different aspects of road safety are critically analyzed. This study highlights important directions for future work and some major challenges such as data collection Poor data quality and lack of ground truth data.