The literature review results suggest that traditional bottleneck identification methods are developed based on performance measures collected from stationary loop detectors (or Bluetooth sensors). However, the applications of such local sensor based methods are usually restricted by the geographical coverage and the density of embedded detectors on the road. In recent years, the coverage and fidelity of vehicle probe data (VPD) have been greatly improved. The possibility of obtaining extensive, continuous, and dynamic VPD from private sectors such as HERE and INRIX offers a great opportunity to identify and assess freeway bottlenecks at the network level. A number of measures of effectiveness (MOEs) can be derived from VPD and be used or bottleneck identification and evaluation, such as the planning time index (PTI), frequency of congestion (FOC), and travel time index (TTI). In this project, the UNCC researchers analyze the feasibility of applying various MOEs to identify and rank freeway bottlenecks. The results indicate that using travel time reliability (TTR) measures (such as FOC or PTI) can reveal only a specific facet of the travel time distribution, but are not be able to quantify the intensity
dimension of the traffic congestion caused by the bottlenecks. As a consequence, a
comprehensive bottleneck identification method which integrates both PTI and TTI is developed.
Since both PTI and TTI are dimensionless travel time-based performance measures and are developed using the same benchmark for each roadway segment (i.e., free-flow travel time), it is reasonable to integrate both measures into the bottleneck identification and ranking framework. By doing so, both dimensions of traffic congestion on each roadway segment can be accounted for. A case study is performed to illustrate the proposed methodology, using a total of approximately 34 million speed records collected in INRIX for four major interstate corridors in Mecklenburg County, NC, in 2015. Freeway bottlenecks are identified and prioritized for a.m., p.m., both a.m. and p.m. peak periods, respectively. The potential causes of each bottleneck group as identified on Mecklenburg interstate freeways are carefully examined by synthesizing the following information: (1) bottleneck identification and ranking results, (2) geometric characteristics around the bottleneck, (3) operational analysis results obtained from the Highway Capacity Software (HCS), and (4) field trip observations. Based on them, a total of 59 scenarios aiming at alleviating bottleneck congestion are designed and evaluated in this project, which include 26 lane-addition scenarios, 15 road pricing scenarios, and 18 combined scenarios (i.e., lane addition and road pricing). Since improved traffic conditions and new infrastructure can directly affect traveler’s route-choice behavior and will lead to a new regional traffic flow pattern, which may either mitigate or exacerbate existing system bottlenecks, a mesoscopic DTA modeling tool is employed by the UNCC team to assess the impact of various candidate bottleneck mitigation strategies on systemwide performances and travel conditions in close vicinity of the bottlenecks. The research findings suggest that under certain conditions, simply adding one more lane at the bottleneck may deteriorate traffic performances. Such counterintuitive results have been widely reported in the literature, and such phenomenon is known as the Braess’s paradox. In addition to that, the researchers also observe the existence of hidden bottlenecks while evaluating candidatebottleneck mitigation projects. Because the causes of bottlenecks can be highly complex and if one is ameliorated, one or more unexpected bottlenecks can quickly emerge downstream. As such, the decision makers must be very careful to ensure that informed decisions are made as to where to apply the bottleneck mitigation countermeasures. Finally, a performance-based framework is developed to assist in assessing and prioritizing candidate bottleneck mitigation alternatives.
The general project ranking framework includes five components: (1) developing candidate bottleneck mitigation projects, (2) evaluating each project, (3) screening of projects, (4) benefit-cost analysis (BCA), and (5) sensitivity analysis. It is envisioned that the proposed framework can provide insightful and objective information for traffic engineers and decision-makers in choosing effective mobility improvement strategies. The second part of this research report presents another bottleneck identification method developed and case study conducted by the NCSU group. The bottleneck identification algorithm uses two thresholds to detect congestion and to filter recurring bottlenecks, where the thresholds are selected based on a robust sensitivity analysis. Three different performance measures are developed to rank and characterize recurring bottlenecks. In this context, the spatiotemporal impact and the probability of bottleneck activation are used as the basis of the performance measures (estimated in miles-hours of congestion). One of the measures called the Daily Impact (DI) of a bottleneck shows the day-to-day variability associated with its impact. The other two measures, namely the Recurring Bottleneck Impact Factor per activation (RBIF) and Overall RBIF account for the impact of a bottleneck over one or multiple days. RBIF per activation is used to conduct the sensitivity analysis for selecting thresholds, and Overall RBIF provides the foundation for ranking the bottlenecks.
The bottleneck identification and ranking algorithm was applied to all interstate facilities within Wake and Durham counties in North Carolina for all weekdays from September to December 2015.
The analysis revealed the presence of 14 recurring bottleneck locations on the study routes along with their activation times. This was based on a temporal threshold activation rate of 33%, along with a threshold congestion speed of 45 mph. Most of the bottlenecks were associated with merging phenomena from nearby on-ramps, and one bottleneck was located at a weaving segment. In addition, two special case bottlenecks are identified, which were activated due to the presence of a long-term work zone (I-40/440 fortify project by NCDOT) and the other due to a queue spilling back from a major arterial (from US-70 to I-440 eastbound at MM 8). The Overall RBIF for the identified bottlenecks vary within a range from 8 to 749 mi.hours of congestion. The distribution of DI for the top-ranked bottleneck (I-40 Eastbound at Wade Avenue/MM 289) showed that its impact can fluctuate significantly (from 0 to 25
mi.hours of congestion). This high variation could be contributed to various non-recurring events occurring within the bottleneck region. Field visits were conducted to verify the activation times of the identified recurring bottlenecks, and 9 out of 14 bottlenecks were verified successfully (note that this ratio is higher than the threshold probability of recurring bottleneck activation of 33%). A video recorder and an in-vehicle tracking device were used to capture traffic conditions as well as the detailed speed profiles of the test vehicle respectively at each bottleneck site. These field visits also verified the contributing factors and demonstrated the bottleneck activations more clearly. To assess the efficiency of a bottleneck mitigation project, a before-after observational study was conducted. It compared the changes in various performance measures such as Overall RBIF, maximum expected queue length and duration of congestion, and probability of activation of a bottleneck before and after implementation of the project. Based on the observed changes in
those MOE’s, the performance of a mitigation strategy was evaluated. To demonstrate, an
improvement project implemented by NCDOT during the period 2009-11 to widen I-40 near
Raleigh was assessed using the proposed method. A target bottleneck was selected within the
spatial extent of the project area. The results indicated that all the (undesirable) performance
measures of the bottleneck decreased in value significantly along with a migration of the
bottleneck further upstream after the implementation of the project.
