A Novel Approach to Detecting Non-recurrent Traffic Flow Patterns in Signalized Arterials

The performance of intelligent transportation systems is dependent on their adaptability to non-recurrent traffic flow events. A novel approach to detecting non-recurrent patterns observed in short-term traffic flow is presented. Traffic flow patterns are distinguished with respect to non determinism and high nonlinearity by analyzing the recurrent behavior of jointly considered series of volume and occupancy; these patterns are indicative of rare traffic events and are further studied. Findings show that rare patterns are likely to occur before traffic flow reaches congestion due to the increased traffic flow complexity. Regarding transitions, traffic flow has a frequent transitional nature, with interesting behavior in synchronized flow; results point to the existence of two regimes with different statistical characteristics with respect to determinism and nonlinearity. The observed statistical characteristics of traffic patterns can be a source of information on the short-term predictability of traffic flow for enhancing the accuracy of traffic forecasting models.