Freeway On-Ramp Bottleneck Activation, Capacity, and the Fundamental Relationship

Almost all freeway delay arises from bottlenecks of one form or another, e.g., incidents, merging, lane drops, and weaving. A bottleneck is said to be active when it is restricting flow. Conventionally an active freeway bottleneck is modeled as if it occurs at a discrete point in space, with queued traffic upstream of the bottleneck location and unqueued traffic downstream. As will be shown herein, this conventional point bottleneck model appears to be too simplistic to capture all of the traffic flow dynamics in the vicinity of bottlenecks. Empirical studies of bottlenecks are encumbered with the challenge of simultaneously estimating capacity and detecting the instant when the arriving demand exceeds the bottleneck capacity. We believe this fact may have helped to obscure a poor fit between the point bottleneck model and actual bottleneck effects that occur over an extended distance. To date a comprehensive understanding of the subtle but important factors that contribute to the bottleneck mechanism remains elusive.;This dissertation examines a merging bottleneck while revisiting commonly held assumptions and uncovering systematic biases that likely have distorted our understanding of bottleneck formation, bottleneck capacity, and even empirical studies of the fundamental relationship (FR) of macroscopic traffic flow. This simulation-based study incorporates microscopic driver behavior with macroscopic traffic flow theory and seeks to provide better insight into those bottleneck features. The simulation extends a conventional car following model to also include a driver relaxation factor for the vehicles that enter or are immediately behind an entering vehicle (termed "affected vehicles"). Rather than instantaneously changing speed, headway or velocity after an entrance maneuver, these affected vehicles initially take a shorter headway and spacing for the given speed than allowed by the car following model and then gradually approach the model's speed-spacing relationship over many seconds.;The simulation results show that the queue initially forms downstream of the on-ramp due to the driver relaxation. The downstream end of the queue then grows further downstream of this formation location, and later recedes upstream back to the on-ramp. The spatial nature of these findings is clearly in contradiction with a point bottleneck model but is consistent with some empirical studies. According to the simulation results, it takes several minutes between the time of the initial bottleneck activation and the time the queue reaches the on-ramp. Simulating conventional detector measurements over the period, we show that flow is higher than the underlying FR would predict (termed "supersaturated flow") in any sample containing an affected vehicle with unsustainably short headway. If one does not already know capacity (as is typically the case in an empirical study) then this systematic bias in flow due to the affected vehicles is not readily apparent in the detector measurements: During the initial queue formation speeds remain close to free speed and the supersaturated states can exceed the bottleneck capacity.;We speculate that the driver relaxation mechanism is common and that many empirical bottleneck capacity studies have erroneously mistaken several minutes of the supersaturated flows to be the bottleneck capacity, when in fact these unsustainably high flows simply reflect the fact that the system is starting to store vehicles further downstream during the earliest portion of bottleneck activation. Instead of flow eventually dropping "from capacity", we see flow drop "to capacity" from supersaturation. This dissertation investigates this issue in detail and examines how the supersaturated states can distort an empirically observed FR.;While this study is based on a simulation work, the major findings provide clues to better interpret past empirical observations that have not been fully explained, yielding new insights into merge bottleneck mechanisms, capacity, and the FR. It is hoped that the work developed herein will help shape future empirical studies to test the validity of the new interpretations and quantify the impacts from the previously overlooked phenomena.