The use of the Cowan M3 distribution for modelling roundabout flow

Abstract

This paper analyses two methods for the estimate of parameters in the M3 distribution. Both methods were derived by Sullivan DP and Troutbeck RJ (see IRRD 866429). The maximum likelihood method was found to give poor results, depending on deviance between the M3 distribution and the observed distribution. The method of moments gave good results. If the minimum headway was fixed the proportion of free vehicles could be related to the traffic flow. These results were used for the analysis of two-lane flow. The M3 distribution, generalised for the two-lane case, was found to fit well to observed data. It also gave good predictions of capacity.