Improvement of an ADT Estimation Model Developed for A-Class Roads in Sri Lanka by Incorporating National Expressways and Important B-Class Links

Abstract

AbstractEstimates of Average Daily Traffic (ADT) are important in the operations of state highway departments for new constructions, improving existing facilities and maintenance. The available manual survey methods to calculate ADT are costly and time consuming. There have been several previous attempts to develop ADT estimation models for the A-class road network in Sri Lanka. One of the serious shortcomings of these models is that the contribution of important class B roads and expressways in transferring inter-district trips is not represented. Hence, this study aims to improve ADT estimation model for the A-class road network including key B-class links and current expressways (E01, E02, E03). Six independent variables were selected in order to represent three main contributors to the ADT through the location; local, regional and inter-district trips, as build-up areas of cities and, distance to the counting stations, population density in the administrative district and the network connectivity factor. Network connectivity factors were calculated based on a link-node system with 96 nodes and 162 links. Road junctions, interchanges of expressways and district capitals are also selected as nodes in the link-node system. Important B-class road links were selected based on average ADT in B-class links in each district, google maps and, local knowledge. Furthermore, B-class links connected with expressway interchanges are also selected based on ADT values. Assuming the rational behaviour of trip makers that try to reduce fuel, toll cost, travel time, travel distance, etc. the network connectivity factor is derived based on a generalized cost function. Generalized cost matrix was used as an input for analyzing the cheapest path using Dijkstra’s Algorithm in Python platform. A regression analysis is done for obtaining the respective parameters of the ADT model with Minitab 19 software. Model resulted a R-square value of 0.71. Percentage bias (PBIAS) was checked for calibration and validation data. For the calibration data, PBIAS is −0.00937% and for the validation data, PBIAS is 1.8%. Hence, the model is not biased and there is a significant improvement of the model, while using generalized cost instead of distance or travel time.KeywordsADTNetwork connectivity factorGeneralized cost functionRegression analysis