Level of service criteria of urban streets in indian context using advanced classification tools

Abstract

In India Level of Service (LOS) is not well defined for urban streets. The analysis procedure followed in India is that developed by HCM 2000. Speed ranges of LOS categories for various urban street classes defined by HCM are appropriate for developed countries having homogenous type of traffic flow. India being a developing country its traffic is very much heterogeneous having vehicles of different operational characteristics. So LOS criteria in Indian context should be defined correctly considering the traffic and geometric characteristics of urban streets. Defining LOS is very important as this is the first step of LOS analysis. LOS analysis is necessitated as this affects planning, design, operational aspects of transportation projects and also allocating resources to the competing projects. Defining LOS is basically a classification problem and application of cluster analysis is found to be a suitable technique to solve the problem. Advanced clustering techniques like Artificial Neural Network (ANN), Affinity Propagation (AP), Genetic Algorithm-Fuzzy(GA-Fuzzy), Partition around Medoid (PAM) can be used to solve the classification problems. Before applying these algorithms for clustering purpose various cluster validation parameters are used to determine the optimal number of cluster for the input data set. This cluster validation parameter holds significance as this decides the numbers of urban street class the free flow speed (FFS) data should be clustered into. After deciding the number of urban street class, the above four algorithms are used in this research in two steps. First, the free flow speed data were clustered to determine the urban street class of each segment. After determination of the urban street class of each segment average travel speed in peak and off peak hour is used in clustering methods to determine the speed ranges of LOS categories. For this study lot of speed data is required for which GPS is found to be the most suitable method of data collection and hence extensively used. Free flow speed and average travel speed during peak and of peak hours and inventory details of road segments are used in
this study. All these data are collected from secondary source for this research work. The FFS speed ranges of different urban street class and travel speed ranges of different LOS categories found from different algorithms are found to be different. So to decide the best clustering algorithm for this study four cluster quality evaluation parameters are used. The best clustering algorithm for this study is decided and the FFS and travel speed ranges resulted from this algorithm is suggested for Indian context. These FFS and travel speed ranges are found to be significantly lower than that mentioned in HCM-2000. Heterogeneous traffic flow and roads having varying geometric and surrounding environmental characteristics are the major reasons for these lower values in FFSs. Presence pedestrians, slow moving vehicles, roadside vendors and on-street parking creates side friction that reduces the travel speed of a commuter. From the clustering analysis of FFS data it can be seen less number of roads in Mumbai are of high speed design (street class-I) or highly congested (street class-IV). More number of road segments is of suburban (street class II) or intermediate (street class III) type. It can be suggested that Greater Mumbai region needs substantial geometric improvements to mitigate the demand of exponentially growing traffic.