Improving the performance of a Bus Rapid Transit (BRT) station using local measurements and microsimulation

Bus Rapid Transit (BRT) is an efficient public transportation system used worldwide. BRT systems are complex networks of bus routes with a myriad of operational parameters that are difficult to optimize. In particular, high occupancy stations often experience bus queueing issues, resulting in reduced performance and user dissatisfaction. Most BRT performance improvement methods ignore bus interactions at docking bays, which limits their applicability to systems where bus queues are common. In this paper, we introduce a local method based on cellular automata aimed at enhancing performance at a single BRT station. The proposed method aims to find the docking bay assignments that both minimize the bus delays and improve the passenger experience at the station. Our method was tested with real data from the Calle 100 station of the Transmilenio BRT system in Bogotá, Colombia. We show that a reduction of 55% in the average delay per bus can be achieved by modifying the docking bay assignment at the station. Our findings indicate that the bus delays are minimized when last sub-stop is the busiest one and the middle sub-stop is the least occupied. This asymmetry is a direct consequence of bus interactions and can only be reproduced by microsimulation methods. Further improvements in the user experience can be attained by incorporating the line similarity as a second optimization objective. This score evaluates the similarity between the next stops of the routes assigned to the same sub-stop. Maximizing this score enables the grouping of similar lines, facilitating passengers’ journey initiation. With the inclusion of this factor in the decision-making process, a Pareto front is proposed to yield a set of optimal solutions from which the system’s operator can choose, depending on their interest. Notably, even when the line similarity score is maximized in the Pareto front, the average bus delay is still reduced by 42% compared to the implemented docking bay configuration. A sensitivity analysis reveals that the optimal solutions are effectively modified by changes in the coefficient of variation of both dwell times and headways, however, the proposed optimal solutions are still close to the optimal and in all cases correspond to significant performance improvements. A better understanding of the distribution of dwell times and headways for all lines allows a closer approximation to the optimal solutions. The proposed methodology can readily be applied to any busy BRT station to look for performance improvements.