Bus Bunching Modelling and Control: A Passenger-oriented Approach

Abstract

Bus bunching indicated the situation when two originally equally spaced bus services running close to each other as the earlier service runs increasingly late or the later service runs ahead. This phenomenon corresponds to bus services unreliability due to the extra passenger boarding, improper dispatching or travelling irregularity resulting in traffic congestions, for example. Bunched buses deteriorate passengers travel experience on public transport services that share the infrastructure with other travel modes. This has become a growing concern in big cities where traffic problems are common. It is challenging to improve the bus service reliability as a variety of chaotic sources, such as traffic congestion or unpredictable travel demand, can affect it. In order to understand the main influence factors of bus bunching in Sydney, this thesis first investigates in detail the bus service operations in Sydney and develops three passenger-oriented bus bunching mitigation approaches to improve the bus service reliability. The real picture of bus bunching in Sydney is scrutinised based on the Sydney bus system automatic vehicle location (AVL) field data. By developing the bus bunching distributions based on different criteria, such as stop location, the bus running direction, month, weekday, time and route number, bus bunching is observed to have a higher possibility in occurring during afternoon peak hours on weekdays in March and July. Additionally, buses with an outbound direction are more likely to be bunched. The ten most heavily-used bus routes with the highest number of bus bunching have also been identified using Sydney AVL data. It is observed that the overlapping routes and shared stops result in more bunches due to the interaction between buses. All the above findings have been used as the input for the simulation modelling. Second, we develop a discrete-event micro-simulation model of a bus service system with overlapping routes. The model tracks the number of passenger waiting at each bus stop as well as the amount of rider inside the vehicle of each running bus. The arrival of riders to stops are modelled as a random process while origin and destination of riders are then stored. With the input, as mentioned above, including the bus service schedule, bus stop locations, and the arrival information of riders, we have modelled Sydney's top 10 routes that have the highest number of bunches, assuming that the bus running speed follows the lognormal distribution. The lognormal distribution is determined by a historical bus running speed data fitting and each link's lognormal distribution parameter for running speed generating can be designated by a MATLAB program based on historical field data for each link. The proposed model offers waiting and in-vehicle time of each rider and can replicate the bus bunching situation in Sydney. To mitigate the number of bus bunching and reduce the unreliability of passengers travel time, we propose three control approaches in this study: (i) bus running speed management (ii) pre-planned limited-stop services (stop-skipping) and (iii) real-time boarding restrict approach. The first two approaches are associated with the planning-level control while the third is related to the operation-level bunching control. Bus running speed management method aims at identifying the links with a large number of bunching and equipping the links with the dedicated bus lane to increase the mean bus running speed and to reduce the variability of the bus running speed. This leads to a reduced likelihood of delays, and hence, mitigate the number of bus bunching. This control has considered minimizing the influence of other traffic modes on speed by selecting the least number of critical links in the network. Pre-planned limited-stop (stop-skipping) service applies as an express service in order to reduce the bus total travel time. The limited-stop pattern is rationalized by skipping the low-demand bus stops determined by historical data. This reduces total dwell time and total travel time, and then, the number of bus bunching. As a pre-planned control, passengers are assumed to be informed with all limited-stop services information, thus, none of the passengers would board on the limited-stop service at the low demand stops. Real-time boarding restrict control is a real-time control level method to reduce bus total dwell time by controlling the number of boarding. This approach is triggered when the headway between two buses is less than a pre-defined scheduled headway, in which the delayed bus will subsequently restrict passenger boarding while allow alighting only. Without passenger boarding, a decreasing number of in-vehicle passengers will alight at the following stop, which reduces the bus total running time and corrects the bus trajectory. Extensive numerical experiments have been carried out to analyse the control strategies using the developed model. The analysis results indicate that all of the control approaches achieve a high performance in reducing the number of bunching; bus running speed management and pre-planned stop-skipping methods also perform superior in reducing passenger travelling time; while real-time boarding restrict control, as a remedy method, have a better potential in steadying the passenger mean waiting time.