Modelling bus delay at bus stop

Yueying Huo, Wenquan Li, Jinhua Zhao, Shoulin ZhuTransport 2015 · DOI

Abstract

A bus may be blocked from entering and exiting a stop by other buses and traffic lights. The objective of this paper is to model each type of delay under these phenomena and the overall delay a bus experiences at a stop. Occupy-based delay, transfer block-based delay and block-based delay are defined and modelled. Bus delay at stop is just the sum of these three types of delay. Bus arrival rate, bus service rate, berth number and traffic lights are taken into consideration when modelling delay. Occupy-based delay is modelled with mean waiting time in Queueing theory. Transfer block-based delay and block-based delay are modelled based on standard deviation of waiting time and the probabilities of their occurrences. Two stops in Vancouver, Canada are selected for parameter estimation and model validation. The unknown parameter is estimated as 0.4230 using Ordinary Least Squares (OLS), which indicates that 42.3% of waiting time variation can be attributed to buses being blocked by the buses in front and red light for the selected stops. Model validation shows the average accuracy rate of the proposed model is 75.07% for the selected stops. Bus delay at stop evidently increases when arrival rate is more than 85 buses per hour for the given service time (50 s), ratio of red time to cycle length (0.65) and berth number (2). We also figure out that bus delay at stop evidently increases when service time is more than 60 s for the given arrival rate (54 buses per hour), ratio of red time to cycle length (0.65) and berth number (2). The proposed model can provide a tool for bus stop design and offer the foundation for service quality evaluation of transit.

Introduction

Urban transit efficiency is often compromised at the very point designed for service: the bus stop. For passengers, the time a bus spends at a stop is perceived as boarding and alighting time. However, for transit engineers, this duration is a complex sum of service time and various delays. Buses frequently encounter a "queue for the curb" when berths are occupied, or find themselves trapped in a berth after service is complete due to downstream traffic signals or other buses.

Existing models often oversimplify these interactions, failing to account for the "blockage effect"—a ripple effect where downstream constraints prevent upstream movement. This paper introduces a comprehensive mathematical model to predict bus delays by categorizing them into three distinct components: occupancy, transfer blockages, and exit blockages.

Methodology

The researchers propose that the total bus delay ($D$) is the aggregate of three specific scenarios:

  1. Occupy-based delay ($D_0$): The time a bus waits because all available berths are currently serving other vehicles.
  2. Transfer block-based delay ($D_t$): Additional waiting time incurred by a bus in the queue because the bus currently occupying a berth is prevented from departing.
  3. Block-based delay ($D_b$): The idle time a bus spends in a berth after passenger service is finished, unable to exit due to traffic lights or a leading bus.

To quantify these, the study employs $M/M/s$ Queueing Theory, assuming Poisson arrivals, exponential service times, and a fixed number of berths ($s$).

Schematic of a bus stop queueing system with $s$ berths.

While $D_0$ is derived from standard mean waiting time formulas in queueing theory, the modeling of $D_t$ and $D_b$ is more innovative. These "blockage" delays are modeled using the standard deviation of waiting time and the probability of specific interference events. The state transitions of the system are represented through a birth-death process, allowing for the calculation of steady-state probabilities for various queue lengths.

State transition diagram for the $M/M/s$ queueing model.

A critical parameter, $ heta$, was introduced to represent the fraction of waiting time variation specifically caused by these blockages. This was estimated using Ordinary Least Squares (OLS) regression based on empirical data.

Empirical Calibration and Validation

The model was calibrated using field data from downtown Vancouver, Canada. The researchers focused on two curb-side, near-side stops at the intersections of Granville & Georgia and Granville & Pender.

The dataset included 21 sets for parameter estimation and 29 sets for validation. Key metrics recorded included arrival times, berth entry/exit times, passenger service durations, and traffic signal timings (specifically a 42s red light within a 65s cycle).

Results and Discussion

The study yielded several key insights into the mechanics of transit bottlenecks:

  • The $\theta$ Factor: The estimated value of $\theta = 0.4230$ indicates that approximately 42.3% of the variation in waiting time at these high-volume stops is directly attributable to blockages from traffic lights or leading buses.
  • Model Accuracy: The proposed model achieved an average accuracy rate of 75.07% when validated against real-world observations.
  • Threshold Effects: Delay does not increase linearly with demand. For a standard 2-berth stop, delays spike exponentially once the arrival rate exceeds 85 buses per hour or if the average service time exceeds 60 seconds.
  • Berth Efficiency Paradox: While adding berths reduces occupancy-based delay, it can simultaneously increase blockage delays. More buses in a single line create more opportunities for the lead vehicle to trap those behind it, especially in near-side stop configurations.

Limitations and Future Work

While providing a robust tool for stop design, the model has specific constraints:

  • FIFO Discipline: The model assumes buses cannot overtake one another, which may not apply to all modern stop designs.
  • Mixed Traffic: Interference from private vehicles or bicycles in non-dedicated lanes is not currently factored into the delay equations.
  • Stochastic Distributions: Future iterations could explore log-normal distributions for service times to better reflect varied passenger behaviors and payment systems.
  • Entry Blockage: The model does not yet account for "rear-end blockages," where a bus in a queue cannot enter a vacant front berth because a bus in a rear berth is obstructing the path.

Summary updated

Abstract

A bus may be blocked from entering and exiting a stop by other buses and traffic lights. The objective of this paper is to model each type of delay under these phenomena and the overall delay a bus experiences at a stop. Occupy-based delay, transfer block-based delay and block-based delay are defined and modelled. Bus delay at stop is just the sum of these three types of delay. Bus arrival rate, bus service rate, berth number and traffic lights are taken into consideration when modelling delay. Occupy-based delay is modelled with mean waiting time in Queueing theory. Transfer block-based delay and block-based delay are modelled based on standard deviation of waiting time and the probabilities of their occurrences. Two stops in Vancouver, Canada are selected for parameter estimation and model validation. The unknown parameter is estimated as 0.4230 using Ordinary Least Squares (OLS), which indicates that 42.3% of waiting time variation can be attributed to buses being blocked by the buses in front and red light for the selected stops. Model validation shows the average accuracy rate of the proposed model is 75.07% for the selected stops. Bus delay at stop evidently increases when arrival rate is more than 85 buses per hour for the given service time (50 s), ratio of red time to cycle length (0.65) and berth number (2). We also figure out that bus delay at stop evidently increases when service time is more than 60 s for the given arrival rate (54 buses per hour), ratio of red time to cycle length (0.65) and berth number (2). The proposed model can provide a tool for bus stop design and offer the foundation for service quality evaluation of transit.
BibTeX
@inproceedings{huo2015modelling,
  author = {Yueying Huo and Wenquan Li and Jinhua Zhao and Shoulin Zhu},
  title = {{Modelling bus delay at bus stop}},
  booktitle = {Transport},
  year = {2015},
  doi = {10.3846/16484142.2014.1003324},
}
Figure