Study on The Factors Governing The Travel Time Reliability of Public Bus Transport System

Abstract

Travel time reliability is the key aspect that indicates the quality of urban public transit service. The reliability is the most preferred parameter by the passengers to decide whether or not to choose the public transit mode of transport. Several factors can affect travel time reliability of the public transit system and it is necessary to understand the impact of these factors on travel time reliability of public transit system. Hence, the present research work aims to study the factors governing travel time reliability of the public bus transport system. Mysore is one of the largest cities in Karnataka state whose transit system has been considered in the present research work, since it generates the Automatic Vehicle Location (AVL) data through the Intelligent Transport System (ITS) infrastructure for public transit. Data collected for the research work comprises of AVL data from Mysore ITS and side friction data collected from study sections using videography method. Mysore city transit vehicles are equipped with the GPS units which provide the Automatic Vehicle Location (AVL) and other trip details of the respective buses. This AVL data has been used to extract travel time of bus routes and segments. The field data extracted from videography includes, side friction elements, traffic volume, and travel time of public bus transit at two different road sections (divided and undivided) during weekdays and weekends. This data is utilised in studying the impact of side friction on travel time reliability of the public transit system. Roadside friction is one of the critical factors which hinders the movement of traffic. The impact of different types of friction elements on travel time depends on their static and dynamic characteristics, as well as the position of friction elements on the carriageway. The data collected at two side friction locations of Mysore city has been used to analyse the impact of side friction on travel time reliability of public transit system. The data have been categorized as static and dynamic side frictions. An approach has been proposed to represent different types of side friction elements with a single index called Side Friction Index (SFI) using relative weight analysis. Travel time reliability is represented using measures such as Buffer Time Index (BTI), Planning Time Index (PTI), Travel Time Index (TTI) and Reliable Buffer Index (RBI). The impact of side friction on travel time reliability was found to be sensitive to traffic volume, and hence the thresholds for different traffic volume levels have been ii determined using K-means clustering method. The impact of side friction on reliability measures at different traffic volume levels has been studied and found to have a non- linear (exponential) relationship. The impact of SFI has been observed to be higher on TTI and PTI in comparison with BTI. The outcomes from this study show that the impact of side friction on TTI and PTI is sensitive to traffic volume, especially at higher traffic volume level and impact of side friction on BTI is least at medium traffic volume level. The inference from this research work shows that the impact of side friction elements varies with respect to the type of road (divided and undivided), traffic volume levels, different days of week (weekday and weekend), and different time periods of day. Travel time variability (TTV) plays a significant role in analysing the reliability of the public transit system. Therefore, this study attempts to analyse travel time variability of the public transit system with the help of AVL data of buses collected from the Mysore ITS. The travel time data are analysed at different temporal aggregation levels corresponding to different Departure Time Windows for peak and off-peak periods. Travel time variability is also influenced by the presence of intersections, bus stops and other geometric and traffic characteristics. Hence, the segment level analysis has been carried out taking into consider the presence of bus stops, intersections and land-use type. AVL data collected from Mysore ITS are used to evaluate travel time distributions with respect to temporal aggregations (peak period, off-peak period, 60 minutes, 30 minutes and 15 minutes) and spatial aggregations (route level and segment level). The distribution fitting process has been carried out using EasyFit software, which estimates the distribution parameters using maximum likelihood estimation (MLE) method. The Kolmogorov-Smirnov (KS) test for goodness of fit has been used to evaluate the fitting of each distribution. The performance of each selected distribution has been evaluated in terms of accuracy and robustness. The results of both route and segment level analysis show that the Generalised Extreme Value (GEV) distribution is superior in describing travel time variability of public transit. The accuracy and robustness of GEV distribution are higher than that of other distributions and also the performance of GEV distribution in the case of signalised intersections and land use type shows the fitting ability and versatility of GEV distribution. Hence, GEV iii distribution has been considered as the descriptor of travel time variability of the public transit system. Travel time reliability measures, TTI, PTI and BTI of four bus routes are determined using GEV distribution and reliability of these routes have been evaluated. The reliability measures of the study routes indicate that the reliability of public transit is lower during peak hours. Understanding the factors causing unreliability of the public transit system is necessary for the improvement of system’s reliability. In this study, the reliability of the system has been modelled considering three travel time reliability measures. The Multiple Linear Regression (MLR) method has been adopted to model the three travel time reliability measures (Average Travel Time (ATT), Planning Time (PT) and Buffer Time (BT)) as the dependent variables and independent variables selected are corresponding to five important factors affecting the measures related to travel time: segment length, bus stops, intersections, land-use and peak/off-peak time period. The results of this study show that length of the segment has a higher impact on all the three reliability measures. The average delay has a higher standardised coefficient value than standard deviation (SD) of delay in the case of ATT and PT. In BT model, SD of delay is more than average delay, which shows that variation in bus stop delay leads to a higher buffer time. The presence of intersection in the segments and Central Business District (CBD)/commercial land-use segments are found to have lesser travel time reliability. Level of service (LOS) is a quantitative stratification of a performance measure or a measure that represents the quality of service. The LOS of bus routes are determined based on travel time reliability such as TTI, PTI and BTI. K-means clustering method has been applied to the segment level travel time data of four bus routes to determine LOS thresholds. Initially, globally accepted six clusters for LOS (A to F) have been considered and cluster validation has been conducted using silhouette analysis. The results of cluster validation show that clusters have reasonable structures and six clusters can be used to determine the LOS thresholds based on these reliability measures. Finally, recommendations have been put forward based on the outcomes of the research work to improve the reliability of the public transit system.