Speed distribution for interrupted flow facility under mixed traffic

Modeling of traffic parameters is a fundamental aspect of solving various traffic and transportation problems. Among them, the speed of the vehicle is often used to evaluate different traffic stream parameters. A detailed investigation of vehicular speed is extremely required in modeling various stream parameters both in theoretical and simulation-based approaches. Speed is a continuous random variable, and most of the studies proposed a normal distribution to describe speed distribution characteristics. However, in a mixed traffic stream, the speed distribution deviates from the normal distribution due to the heterogeneity. Therefore, an extensive investigation of speed distribution characteristics is carried out at signalized intersection under mixed traffic scenarios. A total number of 16 four-legged signalized intersections were selected from four different metro cities (Delhi, Kolkata, Bhubaneswar, and Jaipur) for detailed analysis of speed distribution. Six different distribution models, namely normal, log-normal, gamma, beta, burr, and generalized extreme value (GEV) distributions, are considered to find out the speed distribution model. It is observed that burr and GEV distributions are found most suitable to model empirical speed distribution for mixed traffic environment. Whereas, the GEV distribution shows above 90% fit percentage to describe the speed distribution for each location and each vehicle type. Besides, the normal and log-normal distributions are found least appropriate. Therefore, novel ranking method is proposed as per the goodness of fit result obtained by the Kolmogorov-Smirnov (K-S) statistical test. It is found that the GEV is proven to be appropriate fitting distribution among the other selected distribution model to represent speed distribution for each vehicle type as well for any vehicle composition level. Thus, the present study provides new statistics of speed distribution model that would be beneficial for researchers and engineers to choose most suitable distribution model in modeling vehicular speed at signalized intersections. (C) 2021 Elsevier B.V. All rights reserved.