Lane-Based Breakdown Identification at Diverge Sections for Breakdown Probability Modeling

Abstract

When the breakdown probability on expressways is being modeled, the identification of breakdowns is a decisive issue. To date, most identification methods have treated all lanes of the main line as one unit; these methods have been conducted mainly by checking the aggregated data of the cross section. However, this cross section–based method oversimplifies breakdown identification for expressway facilities, such as diverge sections, on which preferences for lane usage differ significantly. Therefore, a lane-based method is proposed to identify breakdowns in each lane; the timing of the breakdown is determined by a critical speed, which is optimized by obtaining the most significant speed reductions that accompany the breakdown. The proposed lane-based method is applied to four bottlenecks at diverge sections on intercity expressways in Japan. The superiorities of the lane-based method are as follows. First, the method can identify and exclude semicongested cases, in which some lanes are congested and others are not. At a diverge section of the Toyota junction in Japan, 15 semicongested cases were identified by the lane-based method from the 198 breakdowns identified by the cross section–based method. Second, the timing of a breakdown can be appropriately determined, as it was for 34 cases on this diverge section. These superiorities significantly improve the accuracy of extracted breakdown flow rates, which are underestimated by the existing cross section–based method. The breakdown probability is modeled with the lane-based method on the extracted breakdown flow rates, and the impacts of traffic condition characteristics are discussed. The diverge rate and the lane utilization rate are found to have significant impacts on the breakdown probability at diverge sections.