Spatiotemporal capacity estimation of bus rapid transit system based on dwell time analysis

Abstract

The performance study of an urban transport system, particularly a Bus Rapid Transit System (BRTS), must report on its operations and reliability. Such study of BRTS comprises numerous facets, including capacity, which directly influences how the system practically operates and serves the commuters. Hubballi-Dharwad Bus Rapid Transit System (HDBRTS) has been operational since 2018. A performance study is necessary to evaluate the performance of HDBRTS, which aids in its upgradation and improvement. The current research uses the experimental technique through an innovative and inspired basis to comment on the HDBRTS's performance by estimating the corridor's operational capacity at different spatial and temporal fluctuations. The selected route of the HDBRTS comprises combined segregated (exclusive traffic environment) and unsegregated (mixed traffic environment) stretches. The current study mainly conducted a video graphics-based survey to acquire the necessary data on identified spatial and temporal trends at various HDBRTS bus stations. The essential data gathered consists of Dwell Time (DT)-based data at each station, summarising the total time a bus takes to serve passengers at a station. DT is inversely proportional to the capacity of the particular bus station, which is related to the Failure Rate (FR). FR values of all the bus stations of the route were analyzed using DT, and then capacity values were calculated at different spatiotemporal patterns. Study results show that the busiest stations of the identified routes with critical DT values have FR values in the range of 1–2%, contradicting previous studies. The variations in the capacity of the stations, both spatially and temporally, were graphically represented with the minimum capacity of the segregated stretch as 36 buses/hr and the unsegregated stretch as 31 buses/hr. Finally, the Level Of Service (LOS) of the chosen study corridor was developed using the K-Means clustering algorithm and validated using the Silhouette Coefficient technique. The silhouette coefficient values obtained range from 0.52 to 0.74, indicating a reasonable structure.