Transportation Research Part E: Logistics and Transportation Review
A new day-to-day dynamic network vulnerability analysis approach with Weibit-based route adjustment process
Introduction
- (1)
Transportation network vulnerability analysis
Transportation system is one of the important lifelines that are vital to people’s health, safety, comfort, and economic activities. In light of the recent man-made and natural disasters frequently occurred in the world, disruptions to the transportation system can seriously damage the economic productivity of the society and cause inconvenience to peoples’ daily life. These man-made and natural disasters have emphasized the multi-faceted importance and vulnerability of the transportation networks to the society, and the need for government agencies and communities to prepare strategies to identify and strengthen the weakness of transportation networks to withstand losses caused by these disruptive events.
Vulnerability analysis is a powerful tool in identifying the critical components in transportation networks. Among the abundant studies on transportation network vulnerability analysis, the two widely cited definitions of vulnerability come from Berdica, 2002, D'Este and Taylor, 2003. Berdica (2002) defined vulnerability as the susceptibility of a system to threats and incidents that result in operational degradation, while D'Este and Taylor (2003) used the notion of accessibility to define that a network link is vulnerable if the loss of the link significantly diminishes the accessibility of the network. According to Taylor (2017), the main task of vulnerability analysis is to determine the network components (e.g., nodes and links), whose failures or degradations could significantly affect travelers’ behaviors and network performance. Therefore, vulnerability analysis is also known as the identification of the weakest/critical/important/vulnerable components. This identification task has wide applications in both the pre-disaster planning and post-disaster management (e.g., targeted protection or retrofitting, strategic location of rapid response and repair stations, evacuation route planning, and evacuation network monitoring) to ensure that the critical components of the network are adequately monitored and protected.
Transportation network vulnerability analysis has received a great deal of attention in the past decade. Different perspectives, measures, and evaluation methods have been proposed to study the transportation network vulnerability (see, e.g., Chen et al., 2007a, Chen et al., 2007b, Murray and Grubesic, 2007, Nagurney and Qiang, 2010, Luathep et al., 2011, Chen et al., 2012, Ho et al., 2013, El-Rashidy and Grant-Muller, 2014, Gedik et al., 2014, Jenelius and Mattsson, 2015, Reggiani et al., 2015, Oliveira et al., 2016, Wang et al., 2016a, Wang et al., 2016b, Bell et al., 2017, Calatayud et al., 2017, Bababeik et al., 2018, Haghighi et al., 2018, Garcia-Palomares et al., 2018, Xu et al., 2018b, Gu et al., 2020). As reviewed by Mattsson and Jenelius, 2015, Xu et al., 2018b, most existing methods of transportation network vulnerability analysis are based on enumerating/generating/simulating disruption scenarios and then evaluating the network performance based on some performance measures (e.g., origin–destination (O-D) connectivity, total system cost, efficiency, and accessibility). The most widely used model in vulnerability analysis is the classical (static) user equilibrium (UE) traffic assignment model. To the best knowledge of the authors, very few studies used within-day or day-to-day dynamic traffic assignment models in the network vulnerability analysis. Jenelius, 2007, Cats and Jenelius, 2014 considered the dynamics and information effects in the vulnerability analysis of road networks and transit networks, respectively. The “dynamics” considered in their studies belong to the within-day or short-term dynamics, which may be more applicable to the disruptions whose impact period and the subsequent recovery process are within a few hours. However, many disruptions in the transportation network are usually associated with severe infrastructure damages and a relatively long-term reconstruction process (from a few weeks to a few months). Such disruptions and reconstructions will cause the travelers’ day-to-day route or mode choice adjustment afterwards. For example, on May 23, 2016, a truck accident damaged a four-lane segment in the Middle-Ring elevated expressway of Shanghai and also blocked its corresponding surface roads. The reconstruction lasted 14 days. This expressway segment accommodated daily traffic of more than 61,000 vehicles, whose route choices had to be adjusted after this truck accident (Tian and Chen, 2019). Empirical studies have also revealed that the traffic flow conditions of many elevated expressways and surface roads fluctuated dramatically during these 14 days and also afterwards (e.g., Li, 2018, Wang, 2019). Hence, it is important and necessary to explicitly model the day-to-day route choice adjustment in the transportation network vulnerability analysis.
Other than the characteristics of the considered disruption mentioned above, the widely-used static UE traffic assignment has the following drawbacks in the transportation network vulnerability analysis: (1) The existence or attainability of this theoretically ideal UE state has been questioned by many studies (e.g., Iida et al., 1992, Friesz et al., 1994). Instead, the network flows evolve from one realizable disequilibrium state to another; and (2) Managers may activate some recovery strategies far earlier than the long-term “final equilibrium” state after the disruption, which means that it is more meaningful to understand the evolution process rather than just the “final equilibrium” state. Ignoring the post-disaster network flow evolution process may underestimate the negative consequences, and potentially lead to bias or cost-ineffectiveness in the recovery strategy decision-making as well as the inability of differentiating the impact of various recovery schedules.
- (2)
Day-to-day dynamic traffic assignment
As mentioned above, the disruption of a network component is likely to cause fluctuations of the transportation network states (including both supply and demand) and change commuters’ route choice decisions from one day to another. To evaluate the system-wide performance after the disruption, an appropriate day-to-day dynamic traffic assignment model is needed to predict the network flow evolution.
Day-to-day dynamic traffic assignment models have been extensively studied in the literature. They can be classified into different categories according to different dimensions, such as discrete-time models versus continuous-time models, link-based models versus path-based models, and deterministic models versus stochastic models. For a more comprehensive review on day-to-day models, readers may refer to Cantarella and Watling, 2016, Zhou et al., 2017.
In this paper, we classify the current day-to-day models according to two dimensions: route choice criterion and uncertainty consideration (i.e., subjective and objective uncertainty). Herein the subjective uncertainty is referred to as the travelers’ perception error due to their imperfect knowledge about the network states. The objective uncertainty is referred to as the travel time variability that widely exists in transportation systems. In reality, both subjective uncertainty and objective uncertainty affect the travelers’ route choice decisions, especially after significant disruptions. Table 1 provides a selective summary of the existing models.
From Table 1, we can see that the existing studies mainly adopted either deterministic or Logit-based stochastic UE (SUE) route choice models. Also, most of them usually assumed that the link travel times on each day are deterministic without uncertainty. On the one hand, the unrealistic perfect information assumption of the deterministic UE model and the independently and identically distributed (IID) perception error assumption in the Logit-based SUE model, have been widely recognized and criticized by both researchers and practitioners. On the other hand, these models ignore the travel time variability that widely exists in daily travel. Many factors can contribute to the travel time variability, e.g., adverse weather, traffic incidents and accidents, work zone, special events, infrastructure disruption, etc. Empirical studies (e.g., Abdel-Aty et al., 1995) have revealed that travel time variability plays an important role in travelers’ route choice decisions, and it is either the most or second most important factor for most commuters. Very few studies have considered both objective and subjective uncertainties simultaneously in the day-to-day dynamic models.
Table 1 also shows that some limited studies indeed have developed stochastic day-to-day models to incorporate different sources of stochasticity, including the objective uncertainty such as demand or capacity uncertainty, and the subjective uncertainty such as random route choice behavior. Compared to deterministic models that assume the travelers’ route choice mechanism per day is determined in advance, stochastic models consider uncertainty in travelers’ route choice decisions and can provide the probability distribution of flow states (Watling and Cantarella, 2015, Cantarella and Watling, 2016, Hazelton and Parry, 2016). Although stochastic models seem to be more general than deterministic models, the computational burden may hinder their applications in large-scale transportation networks. For example, Cantarella and Watling (2016) proposed a stochastic day-to-day model that considers Logit-based route choice and demand fluctuation, which was solved by Monte Carlo techniques due to the multinomial distribution of route flows. In addition, they still inherited the IID assumption of perception error due to the Logit-based route choice model. Note that the day-to-day dynamic network vulnerability analysis addressed in this study needs to run a day-to-day dynamic model for each generated/enumerated disruption scenario.
Hence, to support the day-to-day dynamic network vulnerability analysis, it is necessary to develop a new day-to-day dynamic model that can reasonably capture the effects of both subjective perception error uncertainty and objective travel time uncertainty on the adjustment of travelers’ route choice, while having computational tractability in large-scale networks.
Motivated by the above observations, this paper develops a new network vulnerability analysis method based on a day-to-day dynamic model under uncertainties. The proposed method includes the following two parts.
Part I- Modeling day-to-day dynamics under uncertainties: We explicitly model two types of uncertainties simultaneously in the day-to-day dynamic model: objective uncertainty due to travel time variability, and subjective uncertainty due to the travelers’ imperfect knowledge about the network states. For the objective uncertainty, we use the mean-excess travel time (METT) concept (Chen and Zhou, 2010) as the travel cost to capture travelers’ risk-averseness against travel time variability. For the subjective uncertainty, we provide a new exploration using an advanced discrete choice model (i.e., the Weibit model with the Weibull perception error distribution) to alleviate the drawbacks of the Logit model. With the Weibit model, we can capture the route-specific perception variance in the route adjustment process, rather than assuming a fixed and constant perception variance among different routes as in the Logit model. Then, the objective uncertainty and subjective uncertainty are integrated into the day-to-day modeling framework. Mathematical properties of the proposed model are also examined.
Part II- Day-to-day dynamic network vulnerability analysis: In evaluating the consequence of disruption scenarios, the existing (static) vulnerability analysis methods based on the static UE or Logit-SUE model only consider the consequence at the final equilibrium state, thus leading to the abovementioned drawbacks. To alleviate these drawbacks, we apply the proposed day-to-day dynamic model to develop a new metric of component importance for network-wide vulnerability analysis. Methodologically, the area circled by the network performance evolution curve derived by the proposed model and the pre-disruption normal network state is quantified. This new metric allows the consideration of both the performance dimension and time dimension, i.e., day-dependent consequence rather than the final equilibrium state. A case study in the realistic Winnipeg network is conducted to demonstrate the advantages of the proposed new metric in identifying the critical bridges.
In summary, the main contributions of this paper are twofold: (1) the development of a day-to-day dynamic model considering both subjective perception error uncertainty (via the Weibit model) and objective travel time uncertainty (via the METT concept) to alleviate the drawbacks of the current day-to-day models that mostly adopt UE or Logit-based route choice criterion as well as ignore the travel time variability, and (2) the development of a component importance metric for day-to-day dynamic network-wide vulnerability analysis by characterizing the post-disruption day-dependent consequence to alleviate the limitations of only considering the final static equilibrium consequence as in the existing studies of vulnerability analysis.
The remainder of this paper is organized as follows. Section 2 presents the day-to-day dynamic model under uncertainties. Section 3 provides a new importance measure for network vulnerability analysis. Section 4 uses numerical examples to demonstrate the features of the proposed model and the new importance measure for network vulnerability analysis. Finally, conclusions and future research directions are summarized in Section 5. The notations and their explanations in this paper are summarized in Table D.1 in the Appendix D.
Section snippets
Modeling day-to-day dynamics under both subjective and objective uncertainties
In this section, we review two classical day-to-day dynamic traffic assignment models and then provide the proposed model under uncertainties.
Day-to-day dynamic network vulnerability analysis
As mentioned in Section 1, most existing studies of vulnerability analysis evaluate the consequence of a disruption scenario based on some performance measures at the final equilibrium state after the disruption. However, many disruptions (especially severe infrastructure damages) encountered in transportation systems have a long-term negative impact and also requires a long period to reconstruct and recover. Commuters’ travel behavior would change after disruptions on a daily basis, and the
Numerical examples
In this section, we use two networks to demonstrate the features of the proposed day-to-day dynamic model and the new metric for network vulnerability analysis. Specifically, network 1 (two-route network) is used to illustrate the advantages of the proposed day-to-day dynamic model. Network 2 (Winnipeg network) is used to demonstrate the property of the proposed metric for network vulnerability analysis.
Concluding remarks
The disruptions encountered in the transportation network vulnerability analysis are generally associated with severe infrastructure damages and last for a long time period (from weeks to months) before the complete recovery. This would inevitably affect travelers’ day-to-day route choice behaviors and result in day-to-day network performance fluctuation. In this paper, we develop a new day-to-day dynamic network vulnerability analysis approach that allows the consideration of day-to-day
CRediT authorship contribution statement
Xiangdong Xu: Conceptualization, Methodology, Software, Investigation, Writing - original draft, Writing - review & editing, Visualization, Validation, Resources. Kai Qu: Software, Investigation, Writing - original draft, Writing - review & editing, Visualization. Anthony Chen: Conceptualization, Methodology, Investigation, Writing - review & editing, Supervision, Resources. Chao Yang: Writing - review & editing.
Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgements
The authors are grateful to five anonymous referees of the 24th International Symposium on Transportation and Traffic Theory (ISTTT 24) for their constructive comments and suggestions to improve the quality and clarity of the paper. This research was supported by the National Natural Science Foundation of China (71971159; 71890973; 72071174), the Research Grants Council of the Hong Kong Special Administrative Region (15267116; 15222221), the Kwong Wah Education Foundation of the Research
References (102)
- et al.
Increasing the resilience level of a vulnerable rail network: The strategy of location and allocation of emergency relief trains
Transp. Res. Part E
(2018) - et al.
Investigating transport network vulnerability by capacity weighted spectral analysis
Transp. Res. Part B
(2017) An introduction to road vulnerability: what has been done, is done and should be done
Transp. Policy
(2002)- et al.
Advanced traveller information systems under recurrent traffic conditions: Network equilibrium and stability
Transp. Res. Part B
(2016) - et al.
Vulnerability of international freight flows to shipping network disruptions: A multiplex network perspective
Transp. Res. Part E
(2017) - et al.
A general stochastic process for day-to-day dynamic traffic assignment: Formulation, asymptotic behaviour, and stability analysis
Transp. Res. Part B
(2016) A Stochastic-process approach to the analysis of temporal dynamics in transportation networks
Transp. Res. Part B
(1989)- et al.
Closed form expressions for choice probabilities in the Weibull case
Transp. Res. Part B
(2008) - et al.
The alpha-reliable mean-excess traffic equilibrium model with stochastic travel times
Transp. Res. Part B
(2010) - et al.
Vulnerability analysis for large-scale and congested road networks with demand uncertainty
Transp. Res. Part A
(2012)
Surrogate-based simulation optimization approach for day-to-day dynamics model calibration with real data
Transp. Res. Part C
A probabilistic multipath traffic assignment model which obviates path enumeration
Transp. Res.
An assessment method for highway network vulnerability
J. Transp. Geogr.
Vulnerability assessment and re-routing of freight trains under disruptions: A coal supply chain network application
Transp. Res. Part E
Link-based day-to-day network traffic dynamics and equilibria
Transp. Res. Part B
A discrete rational adjustment process of link flows in traffic networks
Transp. Res. Part C
Performance of transportation network under perturbations: reliability, vulnerability, and resilience
Transp. Res. Part E
On a link-based day-to-day traffic assignment model
Transp. Res. Part B
Day-to-day variation in markovian traffic assignment models
Transp. Res. Part B
Statistical methods for comparison of day-to-day traffic models
Transp. Res. Part B
A link-based day-to-day traffic assignment model
Transp. Res. Part B
The stability of stochastic equilibrium in a two-link transportation network
Transp. Res. Part B
Experimental-analysis of dynamic route choice behavior
Transp. Res. Part B
Road network vulnerability analysis: Conceptualization, implementation and application
Comput. Environ. Urban Syst.
Unconstrained weibit stochastic user equilibrium model with extensions
Transp. Res. Part B
Robust optimization of distance-based tolls in a network considering stochastic day to day dynamics
Transp. Res. Part C
Network with degradable links: capacity analysis and design
Transp. Res. Part B
Degradable transport network: Travel time budget of travelers with heterogeneous risk aversion
Transp. Res. Part B
Vulnerability and resilience of transport systems - A discussion of recent research
Transp. Res. Part A
Indicators of reliability and vulnerability: Similarities and differences in ranking links of a complex road system
Transp. Res. Part A
Dynamic pricing in discrete time stochastic day-to-day route choice models
Transp. Res. Part B
Transport resilience and vulnerability: The role of connectivity
Transp. Res. Part A
Finding the reliable shortest path with correlated link travel times in signalized traffic networks under uncertainty
Transp. Res. Part E
The existence and calculation of traffic equilibria
Transp. Res. Part B
Dynamic congestion pricing with day-to-day flow evolution and user heterogeneity
Transp. Res. Part C
Travel time unreliability on freeways: why measures based on variance tell only half the story
Transp. Res. Part A
Resilience in transportation systems: a systematic review and future directions
Transport Reviews
Day-to-day congestion pricing and network resilience
Transportmetrica
Identification of critical combination of vulnerable links in transportation networks - a global optimisation approach
Transportmetrica A
Optimal bus service design with limited stop services in a travel corridor
Transp. Res. Part E
Sensitivity analysis based approximation models for day-to-day link flow evolution process
Transp. Res. Part B
Physics of day-to-day network flow dynamics
Transp. Res. Part B
Modeling distribution tail in network performance assessment: A mean-excess total travel time risk measure and analytical estimation method
Transp. Res. Part B
A link-based mean-excess traffic equilibrium model under uncertainty
Transp. Res. Part B
Transportation network redundancy: Complementary measures and computational methods
Transp. Res. Part B
An optimization approach for deriving upper and lower bounds of transportation network vulnerability under simultaneous disruptions of multiple links
Transp. Res. Part C
Day-to-day stationary link flow pattern
Transp. Res. Part B
Bi-objective programming approach for solving the metro timetable optimization problem with dwell time uncertainty
Transp. Res. Part E
Exploration of day-to-day route choice models by a virtual experiment
Transp. Res. Part C
Day-to-day dynamics with advanced traveler information
Transp. Res. Part B
Cited by (18)
Nighttime light remote sensing for urban applications: Progress, challenges, and prospects
2023, ISPRS Journal of Photogrammetry and Remote SensingMeasurement and ranking of important link combinations in the analysis of transportation network vulnerability envelope buffers under multiple-link disruptions
2023, Transportation Research Part B: MethodologicalCitation Excerpt :The range between two bounds can be used to understand the possible variation in network vulnerability with rerouting flows under disruption. To model the rerouting behavior in a specific disruption scenario, one can consider either rational behaviors or irrational behaviors owing to various realistic considerations, including risk-taking behavior due to the objective uncertainty of the disrupted network (Chen et al., 2002a; Chen and Zhou, 2010), perception error due the subjective uncertainty of travelers (Jansuwan and Chen, 2015; Nogal and Honfi, 2019), lack of traffic information under disruptions (Guo and Liu, 2011; Pel et al., 2012; Zheng et al., 2015), and both objective and subjective uncertainty in a day-to-day evolution process (Xu et al., 2021). The modified model M6 is formulated as follows:
A topology-based bounded rationality day-to-day traffic assignment model
2022, Communications in Transportation ResearchCitation Excerpt :Different assumptions about residuals distribution lead to different models, among which the most famous one is certainly the logit model (Watling, 1999; Guo et al., 2013; Smith and Watling, 2016), allowing to find the path choice probability solution in closed form. Other notable variants have been also proposed, such as the C-logit model (Cascetta et al. (1996), the path-size logit and the nested logit (Ben-Akiva and Bierlaire, 2003; Yu et al., 2020), to alleviate some of the weaknesses of logit-based models, as well as the more recent day-to-day weibit-based models (Castillo et al., 2008; Xu et al. 2021). The stationary state of these models leads to a SUE.
Travel time reliability in transportation networks: A review of methodological developments
2022, Transportation Research Part C: Emerging TechnologiesCitation Excerpt :Note that the above models use the common Gumbel variate adopted in Logit-based SUE model (Dial, 1971; Fisk, 1980) as the perception error term to construct the perceived travel time budget. Recently, Xu et al. (2021) develops a Weibit-based mean-excess travel time SUE model in the context of day-to-day dynamics and the perception error follows the Weibull distribution. In the third type, travelers’ perception error is dependent on the actual travel time distribution, and is thus called “stochastic perception error”.
Day-to-day dynamics in a simple traffic network with mixed direct and contrarian route choice behaviors
2022, Physica A: Statistical Mechanics and its ApplicationsCitation Excerpt :Dynamic process (DP) models of route choice provide a flexible framework for predicting the evolution of network states over successive time periods, allowing to explicitly account for the effects of behavioral features such as travelers’ learning, memory, habit and inertia. Building on the seminal work of Horowitz [2], Cantarella & Cascetta [3] defined a unifying framework for discrete-time DP models and established general theoretical results, while several subsequent studies focused on specific issues and applications such as, for example, the identification of attraction domains of traffic equilibria [4], the design and evaluation of Intelligent Transportation Systems [5], the interaction between route choice and traffic-responsive signal control [6,7], and the analysis of commuters’ route choice adjustments following network disruptions [8]. Parallel to the main stream of studies based on discrete-time deterministic dynamical system approaches, continuous-time models of network dynamics have been proposed [9–11], as well as stochastic process models describing network states in terms of probability distributions rather than through point estimates [3,12–14].