Abstract
Abstract. Advances in real-time data collection, data storage and computational systems have led to development of algorithms for transport administrators and engineers that improve traffic safety and reduce cost of road operations. Despite these advances, problems in effectively integrating real-time data acquisition, processing, modelling and road-use strategies at complex intersections and motorways remain. These are related to increasing system performance in identification, analysis, detection and prediction of traffic state in real time. This research develops dynamic models to estimate the probability of road incidents, such as crashes and conflicts, and incident-prone conditions based on real-time data. The models support integration of anticipatory information and fee-based road use strategies in traveller information and management. Development includes macroscopic/microscopic probabilistic models, neural networks, and vector autoregressions tested via machine vision at EU and US sites.
Highlights
Transportation and traffic safety has been identified by academics, local authorities and professionals as a critical problem in urban and regional environments
Advances in real-time data collection systems, data storage, and computational systems have led to algorithms and integrated systems that can be applied in transportation administration and traffic engineering to improve traffic safety and reduce internal and external cost of transport and traffic operations (Maibach et al, 2008; McDonald and Stephanedes, 2002)
The algorithm proposed in this paper is developed to estimate, in real time, future probabilities of incidents, near-incidents and incident-prone conditions based on the flow state, while maintaining efficient data storage
Summary
Transportation and traffic safety has been identified by academics, local authorities and professionals as a critical problem in urban and regional environments. Sensors need to monitor large networks around the clock to collect and manage data to be filtered for reduction to potentially useful data sets. Addressing this need, researchers have defined a modelling framework in which crashes and conflicts appear as subsets of a population of events. They derive relationships between the two subsets, focusing on identifying conditions where data on non-crash events carry information about crashes (Davis, 2003). Our dynamic models estimate the probability of occurrence of road incidents, such as crashes and conflicts, and incident-prone traffic conditions based on real time data. The models are based on results from our research at the EU, NSF, and USDOT, and detailed dynamic data available from EU and US localities
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
