A Comparative Analysis of Exponential and Linear Roundabout Capacity Models Using HCM Research Data

Abstract

This technical note and conference presentation give the results of an investigation to assess basic exponential (non-linear) and linear model forms used in practice for roundabout capacity estimation. Best fit regression and anchored regression analyses were carried out using the HCM single-lane roundabout capacity research data. The full dataset as well as the Glens Falls and Carmel data subsets were analysed. These two data subsets for different roundabout geometry types represent a horizontal slicing of data with different capacity levels over the same range of circulating flows. To assess the applicability of basic exponential and linear models to low and high circulating flow ranges, additional data subsets were used by vertical slicing of the HCM capacity data. Using this method, two-segment linear and exponential models were analysed. Two-segment analyses were carried out using aggregate data as well. The models assessed also included the TRL-Kimber linear capacity model and the HCM exponential model with a new simplified version of the SIDRA geometry method added. The report includes discussions on frequency of data points in low, medium and high circulating flow ranges, the sum of entering flow (capacity) and circulating flow, the ratio of entering flow to circulating flow, issues related to anchored regression models, follow-up headways implied by best fit regression models being larger than measured values, and the quadratic model. As an issue related to underestimation of capacity at low circulating flows, a detailed single-lane roundabout example is given for unbalanced flow conditions under high demand levels. In conclusion, the assessments from various perspectives reported in this document demonstrate the non-linear characteristic of roundabout capacity data as a function of the circulating flow, and support the HCM exponential (non-linear) roundabout capacity model over the linear model form.