ABSTRACT

Data for 44 days from five extended freeway sections around bottlenecks in the San Diego, California, area were analyzed to determine the stability of the point of initial flow breakdown and the feasibility of using similar data for more extensive research into the stability of bottleneck flow phenomena. The ultimate goal of such research is to shed light on the nature of transitions from uncongested to congested flow. Analysis of speed drop sequences suggests that there is rarely a single bottleneck location within critical freeway sections. This in turn suggests that many bottlenecks should be thought of as extended sections rather than points or isolated segments. This suggests an understanding of flow transitions intermediate between the conventional view that flow breaks down consistently at a few locations and the view that flow breakdown is spontaneous and that congested flow is self-organized. Data similar to those used in this study are adequate, but not ideal, for further investigation of the stability of bottleneck flow phenomena. Specific limitations relate to the locations of detector stations and the presence of chronic data errors. This approach to the study of bottlenecks can be improved by combining direct observation with analysis of loop detector data and by using cumulative flow counts to estimate changes in the numbers of vehicles stored in freeway segments.

A complex bottleneck is a section that has several bottlenecks with varying patterns of activation in close proximity to one another; issues of concern are the number and location of individual bottlenecks and the events associated with their activation. The usefulness of N-curve methodology for analyzing complex freeway bottlenecks was evaluated by applying it in the reanalysis of a section of southbound I-5 near San Diego, California, that had been used in a previous study. The reanalysis was based on ordinary loop detector data and considered the relationship between changes in vehicular storage in individual sections (determined by N-curve analysis) and the time series of speeds. Information provided by the N-curve analysis contributed to improved understanding of the section by establishing the existence of multiple bottlenecks in two of the sections, but its contribution was modest when compared with what had already been learned from a careful analysis of speeds. A major difficulty in applying the N-curve technique was the uncertainty of the correction of accumulated errors resulting from count biases. A simple uniform correction proved inadequate in several cases because the biases appeared to be time-dependent; other approaches to bias correction led to plausible results in some of these cases but not others, and in all cases the uncertainty of the bias corrections leads to considerable uncertainty in the queue-size estimates.

No abstract

No abstract