Integrated design framework for on-demand transit system based on spatiotemporal mobility patterns
Introduction
On-demand transit is an intermediate type of mobility service that bridges the gap between collective public transportation and taxi services. By pooling and adaptive routing based on users’ demand, on-demand transit provides more user-centric services than scheduled buses and subways, while providing more collective service than taxis (Daganzo and Ouyang, 2019, Sanaullah et al., 2021). Initially, the concept of on-demand transit was introduced for suburban areas where conventional public transit with fixed schedule cannot be sustained due to scattered and irregular demand patterns (Davison et al., 2014, Li et al., 2022). On-demand transit offers advantages such as flexible schedules and routes, making it a complement and alternative to traditional public transit with appropriate service capacity (Jiang et al., 2018). In recent years, as city dwellers increasingly desire new travel options with high accessibility and reasonable prices (Guo et al., 2017, Noulas et al., 2018, Salnikov et al., 2015), on-demand transit has emerged as a promising solution for urban transportation challenges (Fielbaum and Alonso-Mora, 2020). In particular, on-demand transit is expected to reduce congestion and pollution, while boosting ridership in urban areas (Shaheen and Bouzaghrane, 2019, Hazan et al., 2019).
The on-demand transit operates in accordance with the characteristics of service areas, including physical and socioeconomic environments, as well as demand patterns (Markov et al., 2021, Levin, 2017, Liu et al., 2017). In this regard, Wang et al., 2015, Liu et al., 2017, Brake et al., 2004, Daniels and Mulley, 2012, Errico et al., 2013 have examined past service cases and evaluated their service designs and operational management strategies. Service design includes determining service regions (Rosenberg and Esnard, 2008), service types (Li and Quadrifoglio, 2010, Papanikolaou and Basbas, 2021, Quadrifoglio and Li, 2009), and the size of the service area (Li and Quadrifoglio, 2009, Kim et al., 2019). Operational management is focused on determining the optimal service routes (Bruni et al., 2014), fleet size (Fagnant and Kockelman, 2018, Markov et al., 2021), vehicle capacity (Markov et al., 2021, Bruni et al., 2014), and service schedule (Kim et al., 2019, Fagnant and Kockelman, 2018, Sultana et al., 2018). The level of flexibility is an important factor that characterizes a service, and it can be measured by several criteria, including whether the service has a predefined route schedule, the extent to which the schedule can be changed, and whether the service accepts real-time requests (Errico et al., 2013). Li and Quadrifoglio, 2010, Quadrifoglio and Li, 2009, Papanikolaou and Basbas, 2021 have proposed methods to decide the level of flexibility of a transit service based on the demand density. Specifically, Li and Quadrifoglio, 2010, Quadrifoglio and Li, 2009 have defined critical density and suggested a design methodology for a highly flexible transit that can be applied to regions where demand density is lower than the critical density.
Recent advancements in information and communication technologies have increased the flexibility of on-demand transit by enabling real-time communication (Inturri et al., 2018, Shaheen and Chan, 2016). As a result, there has been growing research on dynamic vehicle routing problems, which aims to find optimal routes based on dynamic requests (Psaraftis, 1988). This topic has been extensively studied in both theoretical and empirical researches (Savelsbergh and Sol, 1998, Berbeglia et al., 2010, Pillac et al., 2013, Psaraftis et al., 2016). When requests are received in real-time, the matching rate may be lowered than when all request information is provided in advance. To increase the real-time matching rate, Tirado and Hvattum, 2017, Srour et al., 2018, Györgyi and Kis, 2019, Ferrucci and Bock, 2016, Goodson et al., 2013, Goodson et al., 2016, Bent and Van Hentenryck, 2004 have addressed dynamic and stochastic problems for the cases when stochastic request information is available. Improving matching rate can increase agency profit and ridership (Tachet et al., 2017), while decreasing congestion and pollution (Alonso-González et al., 2020).
Nevertheless, it is important to note that real-time routing in on-demand transit can result in decreased passenger convenience, especially in case of pooled rides (Lu, 2020, Gurumurthy and Kockelman, 2018, Wang et al., 2018, Moreno et al., 2018). Two main factors contribute to this issue: time variability and extra detours. Time variability can increase due to unexpected itinerary changes resulting from real-time operation, leading to an unreliable service for passengers (Fielbaum and Alonso-Mora, 2020). Several studies have investigated how real-time requests and traffic conditions affect passengers’ journeys (Liu et al., 2019b, Bansal et al., 2019). Extra detours can increase passengers’ ride and waiting time. To address this issue, Fagnant and Kockelman, 2018, Vosooghi et al., 2019, Simonetto et al., 2019 have proposed passenger–vehicle assignment algorithms that improve not only matching rate but also passenger travel time. Fielbaum and Alonso-Mora (2020) suggested a method to control the two factors by utilizing predicted journey information. Furthermore, passenger convenience can be largely improved by integrating transit services with different levels of flexibility (Liu et al., 2019a, Stiglic et al., 2018, Narayan et al., 2020).
Integrating transit services in urban areas is particularly advantageous due to the varying types of demand. Transit services with low flexibility, such as buses and subways, have the ability to efficiently cater to regular and high demand by following fixed routes that are designed based on historical mobility patterns or through trial-and-error. In particular, in urban areas with high commuting populations, traditional public transit services play a vital role in improving passenger convenience and operation efficiency (Nourbakhsh and Ouyang, 2012, Basu et al., 2018). Nonetheless, the transportation equity remains a significant challenge in such areas due to the irregular and low demand that results in lower ridership. Consequently, there is a need for a transit service that can adapt to the varying demand and complement the traditional public transit system. Luo and Nie, 2019, Liu et al., 2019a, Stiglic et al., 2018, Narayan et al., 2020, Basu et al., 2018 have shown that integrating different transit services has many benefits, including reducing total vehicle miles, passengers waiting time, and traffic congestion.
To further enhance the effectiveness of integrating transit services, it is crucial to consider the interplay between transit services and mobility patterns, which have traditionally been studied separately. Therefore, these two research areas do not provide adequate guidance on determining appropriate types of services for various types of cities. The fundamental questions related to the design of integrated on-demand transit services are the following.
How to identify appropriate types of services for urban areas with highly complicated mobility patterns?
How to design efficient routing algorithms for integrated on-demand transit services?
These questions will improve both passenger convenience and operational efficiency of on-demand transit services simultaneously. Accordingly, we aim to propose a design framework for integrating on-demand transit services that can be customized and applied to various urban areas.
This paper presents a design framework for integrating on-demand transit services with different levels of flexibility into a unified transit system. We propose a classification method for identifying mobility patterns, whereby historical demand data is classified into three types: concentrated, clustered, and scattered, based on the spatiotemporal density of origin-destination-departure time (ODT) pairs. We then design an integrated on-demand transit system based on the identified mobility patterns. The proposed system can provide three transit services: (i) planned-and-inflexible (PI) service; (ii) planned-and-flexible (PF) service; and (iii) unplanned-and-flexible (UF) service. The main contributions of this paper are summarized as follows.
we develop a design framework for an on-demand transit system with multiple levels of flexibility that can cover diverse mobility patterns;
we design a demand classification method for identifying mobility patterns depending on the spatiotemporal distribution of trip data; and
we quantify the gain of the proposed methodology through simulation studies using real-world mobility data.
The rest of this paper is organized as follows. After describing the design framework for on-demand transit system in Section 2, we propose a demand classification method in Section 3. We then present a routing algorithm for the integrated on-demand transit system with different levels of flexibility in Section 4. The performance of the proposed system is evaluated based on simulation studies in Section 5. We provide a discussion and suggest future works in Section 6. Finally, a short summary follows in Section 7.
Section snippets
Service types of the on-demand transit system
Planned-and-inflexible (PI) service is the least flexible service in the proposed on-demand transit system. This service is designed solely based on historical demand data, and there are no updates to the route and schedule in response to received requests. The schedules and routes of service vehicles are optimized using the ODT pairs of the concentrated type that are highly probable to occur within a small spatiotemporal range.
Planned-and-flexible (PF) service has an intermediate level of
Demand classification for on-demand transit system
This section presents a method for classifying mobility data based on spatiotemporal densities of ODT flows and identifying diverse mobility patterns. The classification criteria include the minimum density of expected passengers , the minimum spatial range , the minimum temporal range , the maximum spatial range , and the maximum temporal range , where and . The concentrated type includes ODT pair groups in which the expected number of passengers
Route planning
Let be a set of stops, where the stop is the depot that acts as the starting and ending point for all vehicle’s travels. The travel time from the stop to the stop is defined by a function . The set of concentrated type flows is defined by . The flow in is represented by for . All the origins and destinations of are included in as . The disjoint subset of the flow group is denoted as , and
Data set
This section presents a simulation study to evaluate the performance of the proposed integrated on-demand transit system. A simulation experiment was performed using smart card data from Sejong city in South Korea. Smart card data provides travel information on the movement of public transport users, including boarding time and station, alighting time and station, route ID, fare and user type of each transaction. Since the smart card is a payment method used by many public-transport passengers
Transit service design strategy based on spatiotemporal demand density
As travel patterns in urban area have become increasingly complex over time (Sun and Axhausen, 2016), it has become necessary to design transit services that can handle heterogeneous and time-varying demand patterns, taking into account the interrelationship between mobility patterns and operation strategy of transportation system. This paper proposed a design method to identify suitable service types for various urban areas and adjust the level of flexibility of transit services according to
Conclusions
This paper proposes a novel data-driven framework for designing an on-demand transit system. The proposed system provides PI, PF, and UF services, supporting requests that are matched to the concentrated, clustered, and scattered types of ODT pairs. By integrating transit services with different levels of response to real-time demand, the proposed on-demand transit system improves both passenger convenience and operation efficiency simultaneously. The superiority of the proposed system over the
CRediT authorship contribution statement
Jeongyun Kim: Conceptualization, Methodology, Writing – original draft. Sehyun Tak: Methodology, Data curation. Jinwoo Lee: Methodology, Writing – review & editing. Hwasoo Yeo: Conceptualization, Methodology, Writing – review & editing.
Acknowledgments
This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (2022R1A2C1012380).
References (63)
- et al.
Value of time and reliability for urban pooled on-demand services
Transp. Res. C
(2020) - et al.
Dynamic pickup and delivery problems
European J. Oper. Res.
(2010) - et al.
Demand responsive transport: towards the emergence of a new market segment
J. Transp. Geogr.
(2004) - et al.
Designing robust routes for demand-responsive transport systems
Transp. Res. E Logist. Transp. Rev.
(2014) - et al.
Modeling horizontal and vertical equity in the public transport design problem: A case study
Transp. Res. A
(2019) - et al.
A survey of demand responsive transport in great Britain
Transp. Policy
(2014) - et al.
Using Lorenz curves to assess public transport equity
J. Transp. Geogr.
(2011) - et al.
A survey on planning semi-flexible transit systems: Methodological issues and a unifying framework
Transp. Res. C
(2013) - et al.
Pro-active real-time routing in applications with multiple request patterns
European J. Oper. Res.
(2016) - et al.
Unreliability in ridesharing systems: Measuring changes in users’ times due to new requests
Transp. Res. C
(2020)
Analyzing the dynamic ride-sharing potential for shared autonomous vehicle fleets using cellphone data from Orlando, Florida
Comput. Environ. Urban Syst.
A probabilistic approach to pickup and delivery problems with time window uncertainty
European J. Oper. Res.
Congestion-aware system optimal route choice for shared autonomous vehicles
Transp. Res. C
Feeder transit services: Choosing between fixed and demand responsive policy
Transp. Res. C
A framework to integrate mode choice in the design of mobility-on-demand systems
Transp. Res. C
Dynamic shared autonomous taxi system considering on-time arrival reliability
Transp. Res. C
Impact of ride-pooling on the nature of transit network design
Transp. Res. B
Simulation-based design and analysis of on-demand mobility services
Transp. Res. A
Integrated route choice and assignment model for fixed and flexible public transport systems
Transp. Res. C
A structured flexible transit system for low demand areas
Transp. Res. B
Analytical models for comparing demand responsive transport with bus services in low demand interurban areas
Transp. Lett.
A review of dynamic vehicle routing problems
European J. Oper. Res.
A novel two-level optimization approach for clustered vehicle routing problem
Comput. Ind. Eng.
A methodology to derive the critical demand density for designing and operating feeder transit services
Transp. Res. B
Spatio-temporal analysis of on-demand transit: A case study of belleville, Canada
Transp. Res. A
Real-time city-scale ridesharing via linear assignment problems
Transp. Res. C
Enhancing urban mobility: Integrating ride-sharing and public transit
Comput. Oper. Res.
Modeling frequency of rural demand response transit trips
Transp. Res. A
Understanding urban mobility patterns with a probabilistic tensor factorization framework
Transp. Res. B
Shared autonomous vehicle simulation and service design
Transp. Res. C
Exploring the propensity to travel by demand responsive transport in the rural area of lincolnshire in England
Case Stud. Transp. Policy
Cited by (1)
Design of mixed fixed-flexible bus public transport networks by tracking the paths of on-demand vehicles
2024, Transportation Research Part C: Emerging Technologies