Models of Roundabout Lane Capacity

Abstract

Accurate roundabout capacity models are essential for optimal roundabout designs, but there exist significant differences in the predicted capacities of various state-of-the-art models and in their included explanatory variables. An empirical study into roundabout lane entry capacity was thus performed in the U.K. using data from 35 roundabout entry lanes, where various model forms and explanatory variable sets were tested. Two regression models and an artificial neural network were developed. A negative exponential relationship with circulating flow predicted lane capacity better at high and low circulating flows, and better reflected the overall trends in the aggregated capacity data, compared to a linear model. The regression models performed relatively well and provided better information on the impacts of the variables than the neural network. The models consistently suggest that entry-exit separation and flows exiting on the same arm have stronger significant effects on capacity than variables such as entry angle and entry radius. These findings could thus contribute to an improved understanding of the variables that affect entry lane capacity and therefore the development of better roundabout capacity models.