Next Generation Simulation (NGSIM)
-Improved Simulation of Stop Bar Driver Behavior at Signalized Intersections-
Los Angeles, CA and Atlanta, GA Data Sets

Working Papers

The results of a recent study of the headway and lost time at three single-point urban interchanges (SPUIs) are summarized. The data base, containing more than 38,000 headway observations, was collected primarily in the Tampa, Florida, area. The data were used to calculate the minimum discharge headway and start-up lost time for the SPUI's three basic movements: cross road left-turn, off-ramp left-turn, and cross road through. It was found that traditional procedures for estimating the minimum discharge headway may be biased toward values higher than ultimately achieved by the traffic queue. Moreover, the degree of bias varied widely among the movements and sites studied because of unequal numbers of observations. As a result, initial attempts at a cause-and-effect analysis were clouded by a high degree of variability in the data. In recognition of the aforementioned bias, alternative statistical analysis techniques and regression models were used to identify significant effects and to calibrate predictive models of minimum discharge headway and start-up lost time. The results indicate that the minimum discharge headway of the SPUI's two left-turn movements are significantly lower than its through movements and lower than values traditionally used for protected left-turn movements under "ideal" conditions. In fact, the calibrated models predict minimum discharge headways that are generally lower, and start-up lost times that are higher, than those calculated by traditional procedures. Left-turn headway was also found to vary with turn radius.

This paper appears in Transportation Research Record No. 1365, Highway Capacity and Traffic Flow [Year of Publication 1992].

Figures (4); References (9); Tables (3)

Summary

Introduction

The first Single Point Urban Interchanges (SPUIs) were constructed in the US and Europe in the early 1970s. This paper summarizes a portion of the results of a larger study of the operational efficiency of the SPUI. In particular, it describes statistical analyses of headway and lost-time data at three signalized SPUIs in the Tampa, FL area. The analyses consist of factors affecting discharge headway and also comparison of the SPUI data with similar data collected at two At-Grade Intersections (AGIs; one near to one of the SPUIs and one in Texas). Data was collected using tapeswitches placed perpendicular to the direction of traffic flow (to collect vehicular count data) and photocells connected to load switches in the controller cabinet (to monitor indication changes).

Background

The advantage that SPUIs hold over a more traditional diamond interchange is that it requires at the most one signal to serve all movements, while the latter needs two separate signalized junctions (one for each on/off ramp). This may make vehicles stop twice on their course through the intersection. SPUIs have three basic movements to be considered: cross-road left turn movement, off-ramp left turn movement and cross-road through movement.

The following characteristics of signalized intersections are discussed in the paper:

Discharge Headway: According to the Highway Capacity Manual (HCM), the minimum discharge headway H is achieved after the fifth vehicle passes the stop bar (as vehicles after queue position 5 are assumed to be unaffected by startup lost time). Thus, it is calculated by averaging the headways between vehicles at the fifth queue position onwards.

Saturation Flow Rate: It is calculated as 3600/H for through movements (per hour of green). The HCM also recommends lower saturation flow rates for turning movements depending on lane and/or phasing configuration. For turning movements, Kimber et al. (reference 4 in the list) suggested an inverse parabolic relationship between turning radius and saturation flow rate.

Start-up Lost Time: It is the initial part of the green time that is used inefficiently by the first few (four, according to HCM) vehicles in a dissipating queue. The total start-up lost time is calculated by adding up the individual lost times of the first few vehicles. Hence, its value is directly dependent on the value used for H. The HCM indicates that the start-up lost time is generally 2 seconds per phase.

Analysis Methodology, Results and Conclusions

The mean and standard deviation were quantified for headway and lost time data. The factors affecting these variables were identified using ANOVA and multiple comparison tests.  The effect of influential factors and model parameters were then quantified using least-squares regression techniques.

Contrary to HCM recommendations, the SPUI left-turn headways were found to be smaller than the through movement headways (this trend is confirmed by ANOVA analysis further in the paper). Also, left-turn lost times were found to be larger than those of the through movements. From these two observations, the author inferred a trend towards larger start-up lost times for those movements with smaller minimum headways. This trend was anticipated by them because minimum headway is used in the calculation of start-up lost time, as mentioned before . These ‘less than satisfying’ (contradictory to HCM, etc.) results prompted the author to look for inherent biases used in the calculation of minimum discharge headway. He postulated that the minimum headway may not be achieved by the fifth queue position but may be achieved even further back in the queue. This bias is magnified due to the variability of observations. It affects estimation of start-up lost times, because it depends on the value of H.

To counter this bias, the author then considered 13-vehicle queues for the through movements and 10-vehicle queues for the turning movements, and only used queue positions having 20 or more observations to add stability to the analysis. Using this data, he concluded that HCM procedure always overestimates the minimum discharge headway. This overestimation is greater for the left turning movements than the through movements. Considering this overestimation is important because the true effect of a treatment or a factor may be clouded by variable bias by queue position and different frequencies of observation at each queue position.

Left turn headways were found to inversely vary with turn path radii. More findings include lower left turn headways (already mentioned before) than through movement headways, possibly due to higher state of alertness of left turning driver, their ability to see the leading vehicle make the turn and in turn choose the correct path and turn-movement lane striping in the conflict area. The author also found that the through movement headways for SPUIs are significantly larger than the through movement headways for AGIs, possibly because of extra caution exercised by drivers in the long SPUI conflict area. Increased lane volume per cycle (‘traffic pressure’) was found to reduce headways at all queue positions.

Since many of these findings are contrary to conventional trends, the author advocates further research.