Abstract
This study examines the service network design problem (SNDP) for passenger shuttle buses in the airport and nearby places (e.g. train stations, parking, hotels, shopping areas). A time–space service network for bus flows and time–space networks for passenger flows are developed. Based on proposed time–space networks, the studied SNDP is formulated as a mixed integer linear program (MILP) for a single-type bus fleet and deterministic passenger demand, where the objective is to minimize the weighted sum of passenger cost and service operating cost. We then extend the developed SNDP model to the heterogeneous multi-type bus fleet case and the stochastic demand case. To solve the stochastic demand case, a Monte Carlo simulation-based approach is adopted, which is further coupled with the ‘effective demand’ concept (mean demand value plus a margin). The proposed SNDP models and solution approach are applied on the inter-terminal transport network at Sydney Kingsford Smith Airport for illustration.
Acknowledgments
We would like to thank the anonymous referees for their useful comments, which helped us improve both the technical quality and exposition of this paper.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Notes
1 Autonomous vehicles have received substantial attentions in recent years (Chen et al. Citation2016; Liu Citation2018; Wu et al. Citation2020; Zhang et al. Citation2020; Chen and Li Citation2021) and have great potential to be utilized for fixed-line shuttle services.
2 The passenger volume is obtained from ‘2019 Sydney Airport Full Year Results Release’ that is released on 20 February 2020 (https://www.sydneyairport.com.au/investor/investors-centre/asx-newsroom).
3 The golden-section search method can be used for finding the minimum of the objective function (minimization problem) inside a specified interval of Δ (Kiefer Citation1953). For a strictly uni-modal function, the golden-section search is able to find the minimum (minimization problem).
4 We have conducted extensive numerical experiments in order to identify a proper threshold, where the value of tends to stabilize, i.e. the percentage error between two recent values of
is no greater than 0.1%. In our case study, with 30 runs of simulations the estimation of
stabilizes. To ensure consistency and solution quality, we indeed use 100.
5 These settings are comparable to those suggested by ‘Transport for NSW Economic Parameter Values’ (https://www.transport.nsw.gov.au/news-and-events/reports-and-publications/tfnsw-economic-parameter-values/).
6 The demand is generated according to the following. We first obtain the hourly flight volume in 2019 for each terminal in SKSA. We then estimate the number of passengers arriving at terminals based on the flight volume and capacities of domestic and international flights. We further assume that a certain percentage of these passengers will make a connection trip.
7 This is a daily value converted from the price and life span of reported small autonomous vehicles (https://www.prnewswire.co.uk/news-releases/autonomous-shuttles-idtechex-report-reveals-the-future-of-last-mile-mobility-819532388.html/) (https://techcrunch.com/2019/08/26/ford-says-its-autonomous-cars-will-last-just-four-years/).
8 The value of ad-hoc service operating cost is comparable to the car rental cost (https://gogocharters.com/blog/charter-bus-prices/).
9 For the inter-terminal network at SKSA, currently passengers can only walk for the connection trips between two domestic terminals T2 and T3 (in both directions and it takes 5 min), and can take bus line 400 or 420 for connection trips between terminal T1 and terminal T2 or T3 (e.g. 20 min headway).