Estimating the influence of disruption on highway networks using GPS data

https://doi.org/10.1016/j.eswa.2021.115994Get rights and content

Highlights

  • The method is based on the Bayesian theory and thrice-standard-error principle.

  • The method can better account for the data fluctuation and deduce the results.

  • Disrupted roads, detour roads, and congestion roads can be accurately determined.

  • GPS data make the estimation of disruption more quickly and the coverage wider.

Abstract

Incidents, such as natural disasters, public events, and holidays, often cause problems to highways, even paralyze the operation of the whole networks, leading to a serious threat to travel efficiency and safety of the public. To provide better transport management and plans for emergencies, it is important to quickly and accurately identify such incidents and estimate their disruptive effects on the networks. To this end, a novel approach has been proposed in this paper, which is based on the Bayesian theory and thrice-standard-error principle while utilizing vehicle GPS data. Two important indicators, including traffic flows and congestion indexes, along with their change ratios, are built to detect the incidents and evaluate the extent of the disruption. The specific disrupted and detour roads are further determined. The proposed method has been tested using two real-world events in China, and the potential and effectiveness of this technique are demonstrated. With more and more vehicles being equipped with GPS devices worldwide, the designed method can be easily transferable to other countries, paving a way for the adoption of the approach for a more spatial–temporal sensitive highway network disruption analysis method that supports the establishment of a more resilient transport system for emergencies.

Introduction

Highway networks are the foundation of social progress and economic growth, which provide express access to cities and are of paramount importance to the integration of urban agglomerations and the development of the combined area. Incidents, such as natural disasters (e.g. heavy rain, floods, and landslides), traffic accidents (e.g. vehicle collision and tunnel fire), public events, and holidays, often cause problems to highways, even paralyze the operation of the whole networks, leading to a serious threat to travel efficiency and safety of the public. Thus, it is an important part of transport management to quickly and accurately identify such incidents and estimate their disruptive effects on the networks.

Once an incident has occurred, certain transport lines or areas would be disturbed. Traffic conditions on the corresponding highways would be affected, including road capacity, travel time, congestion levels, traffic flows on certain roads (i.e. the changes in people’s travel routes), etc. Methods have been developed to evaluate the disruptive effects of various incidents on road networks. Several major parameters have been considered, including vulnerability/robustness, accessibility, and traffic flows. The representative works are as the followings. (1) Vulnerability/robustness. Vulnerability and robustness describe road networks from an opposite perspective. The former characterizes the likelihood at which a road network (or a single road) is susceptible to disruption by incidents; while the latter reflects the extent to which a network (or a single road) can maintain the functions originally designed for under all circumstances (including those that deviate from normal conditions). Efforts have been made to examine each of these two parameters, which are exemplified by the following studies. Regarding vulnerability, Chen, Lam, Sumalee, Li, & Li, 2012 as well as El-Rashidy and Grant-Muller (2014) developed indexes to determine the most vulnerable links in an urban road network, under the closure of certain major roads. Similarly, Jenelius & Mattsson, 2012, Jenelius & Mattsson, 2015 adopted Grid Cell Approaches to analyze the vulnerability of road networks under the shutdown of either a single link or a part of the network. Moreover, Pedrozo-Acuna et al. (2017) found the most vulnerable points along highways that were disrupted in a flood. In terms of robustness, Sullivan et al. (2010) considered different capacity-reduction rates (e.g. from 100% to 0% capacity-reduction) for detecting the most important links and quantifying their robustness in three virtual networks constructed with various disruptive scenarios (e.g. earthquakes, floods, and tornados). Moreover, Bagloee et al. (2017) defined a global robustness index for ranking links under the similar scenarios, based on Discrete Network Design Approaches. (2) Accessibility. Accessibility reflects the ease (e.g. travel time or travel costs) of a road network with which desired destinations can be reached. It does not only consider travel efficiency (e.g. travel speeds) on the roads, but also the distribution of activity locations across the network. The major work of adopting this parameter includes the research (Sohn, 2006) that examined by how much the accessibility of certain major roads has decreased due to the disruption of a flood, and the study (Taylor and Susilawati, 2012) that identified the locations with the highest reduction in accessibility under the failure of several important links. (3) Traffic flows. Traffic flows depict the total number of vehicles passing a given point in a given time. This parameter has been a major factor for disruptive effect analysis. The typical works include the followings. He and Liu (2012) studied the changes in traffic flows in the highway network of the Twin Cities before and after the I-35W Bridge collapsed; while Danczyk et al. (2017) further characterized the dynamics of the flows in a cordon zone (i.e. a designated area providing alternative routes to travelers in place of the disrupted area) after this incident. Moreover, Pregnolato et al. (2017) examined the relationship between traffic flows and the depth of standing water on the affected roads during a flood.

From the above literature reviews, it was noted that different methods (and parameters) have been developed to study the potential effects of incidents on road networks. Nevertheless, despite the variety of approaches, they are still subject to a number of limitations. The first lies in the analysis of traffic congestion. Traffic congestion has a great impact on travel efficiency, especially after an incident. The roads damaged by an incident and thus removed from the network operation can lead to increases in traffic flows and congestion levels on the adjacent links (Sullivan et al., 2010). Moreover, while the three major parameters, including vulnerability, robustness, and accessibility, analyze the disruptive effects primarily from the view of physical structures of road networks; the remaining parameter of traffic flows, along with the variable of traffic congestion, examine the impact of incidents from the aspect of observed traffic conditions on the roads. While the former analysis could lead to long-term improvement in the design of road network structures for emergencies, the latter is particularly important for formulating control strategies to effectively manage the real-time traffic flows and road conditions and to optimize the current travel situations during the course of the incidents. However, although the analysis of traffic congestion is of great significance, there is still a lack of research that concentrates on this variable and examines the status of congestion in road networks after incidents.

The second limitation concerns the detection of disrupted and detour roads. The determination of these two types of roads provides important guidance for emergency management and planning. While the majority of the existing research focuses on recognizing the most critical links or locations, e.g. in terms of vulnerability and accessibility (Pedrozo-Acuna et al., 2017, Taylor and Susilawati, 2012), only a few studies examine people’s choices for detours. The latter can be represented by the following works. Sohn (2006) assumed the second shortest-travel-time-path as the detour road if the first one (e.g. the regular route) was disrupted. Danczyk et al. (2017) further revealed an avoidance phenomenon after road collapse: drivers drastically avoided areas near the disruption sites by taking alternative routes, but gradually returned to their normal routes after a period of time when the road conditions were recovered. Balal et al. (2019) defined that the detour route included the off-ramp, the on-ramp, and the local arterials connecting the off-ramp and the on-ramp. Moreover, Jenelius & Mattsson, 2012 recognized that it was even difficult to find detours if the damage by the incidents was too severe. Nevertheless, in the above-described research, User Equilibrium Assignment and GIS-based methods have been used in the search for the possible detours (i.e. the second shortest-travel-time-path). It was based on the assumption that travelers try to minimize travel time between their origins and destinations. However, as indicated by the work (Wolf et al., 2004), this is not necessarily the case. Drivers do not actually choose the shortest routes but more often the ones longer than hypothesized, due to a variety of other reasons, such as congestion levels, travel time reliability, and personal preferences. Thus, a method to accurately identify the disrupted and detour roads that are actually occurring during the incidents is still absent.

The third limitation is with respect to the data. The data utilized in the existing research has been obtained from various sources, including sensors (e.g. loop detectors, videos, and Automatic Traffic Counters) (Danczyk et al., 2017, Villiers et al., 2019), field surveys (Pedrozo-Acuna et al., 2017), traffic volumes estimated by traffic assignment and travel demand models (El-Rashidy and Grant-Muller, 2014, Jenelius & Mattsson, 2015), GIS (Taylor and Susilawati, 2012), and simulation of virtual networks (Sullivan et al., 2010). All the data belongs to the traditional way of data acquisition that has inherent constraints (e.g. low coverages, limited sampling frequencies, and/or low levels of data accuracy) (Liu et al., 2013, Liu et al., 2014), leading to a certain level of deviation between what is revealed by the derived results and what the actual road conditions are during the incidents (Cui et al., 2016a, Cui et al., 2016b).

The advancement of Global Positioning Systems (GPS) has created the opportunity to use this technology as a new travel data collection method. For vehicles that are installed with GPS devices or for the drivers who carry smart phones with GPS features, the accurate travel routes and travel speeds can be monitored automatically, providing detailed spatial–temporal travel information and real-time traffic conditions. The added value of GPS data in transportation has been manifested by a variety of research and applications, for example, the construction of travel demand models and the analysis of traffic conditions (Cui et al., 2016a, Cui et al., 2016b) for the whole population. Particularly, regarding traffic condition analysis, GPS data has been widely applied to build Intelligent Transportation Systems to monitor real-time traffic status (e.g. driving speeds along with links) and predict short-term traffic evolution (e.g. congestion). Comparison has been made between the driving speeds derived from traditional sensor data (e.g. data from loop detectors) and those obtained by means of GPS data. Large differences were found: for highways, the GPS-data-based speeds were on average 6.3 km/h (kilometer per hour) lower than the sensor-based ones; while on low-level urban roads, this decrease reached more than 20 km/h. Owen and Levinson (2012) further stressed that the direct traffic measurement using GPS was more accurate than the traffic parameters that were inferred based on traditional data alongside a set of sophisticated models, and that GPS data was the most promising way for transport analysis, and for the identification of incidents and the estimation of the disruptive effects in particular in this paper.

To extend the current research on the analysis of disruptive effects of incidents, and particularly to address the limitations with respect to the development of approaches for accurately modeling congestion status and determining both disrupted and detour roads, a novel method has been proposed in this study. The approach provides the following advantages. (1) It is based on the Bayesian theory and thrice-standard-error principle to better account for the fluctuation of the observed data and deduce the results more accurately. (2) Two important indicators, including traffic flows and congestion indexes, along with their change ratios, are built to identify incidents and estimate the impact of the disruption. Both disrupted and detour roads, as well as the roads where congestion considerably increases, are also determined. (3) Through highly detailed GPS data recorded from vehicles running in the network, the method enables better modeling the indicators and detecting changes of the variable values with high spatial–temporal sensitivity. This further leads to the identification of incidents and the estimation of the disruptive effects more quickly and precisely. (4) In many countries around the world, GPS devices are already installed in vehicles or on smart phones, generating no extra financial costs in terms of data collection, making it a cost-effective and easily transferable approach. (5) Particularly, in this study, the GPS data collected from large amounts of vehicles operating in the highway network of China has been explored, and two real-world events including the New Year’s Day and a snowstorm are analyzed. The potential and effectiveness of this technique are demonstrated and significant results are obtained. (6) The Bayesian theory has been widely used in transportation science to build traffic models (Chen et al., 2019, Khattak and Fontaine, 2020, Mil and Piantanakulchai, 2018, Wang et al., 2020, Xiao et al., 2020, Yu et al., 2019, Zhu et al., 2019) and analyze traffic conditions (Afrin and Yodo, 2021, Li et al., 2021, Fu and Sayed, 2021, Sun et al., 2016, Zheng et al., 2021, Zheng and Sayed, 2019). However, to the best of our knowledge, no studies have yet adopted the theory to build indicators directly from observed mobility data (e.g. GPS data) for disruptive effect analysis during emergencies. When compared to the literature, the proposed method is relatively simple, but practically acts as an effective approach in emergency analysis in real time. Particularly, the method detects both disrupted and detour roads as well as congestion-increasing roads, providing valuable information in formulating traffic control strategies. From this perspective, the application of the Bayesian theory to the identification of incidents and the analysis of potential disruptive affects in the proposed approach, represents an important contribution to the advancement of this technique in the field of transport emergency studies.

The rest of this paper is organized as follows. Section 2 introduces the Bayesian theory and thrice-standard-error principle. 3 Incident identification, 4 Incident impact analysis detail the process of identifying an incident and estimating the impact of the disruption. Sections 5–8 present a case study using two real-world events, and Section 9 further elaborates on an additional application. Finally, Section 10 ends this paper with major conclusions and discussions.

Section snippets

Basics of the model

Normally, traffic flows in a highway network each day or over the time of the day have a stable trend among days with the same day-of-the-week. However, when an incident occurs, this stability could be broken, based on which incidents can be identified and their effects can be estimated. This can be well illustrated in Fig. 1, which describes the distribution of the total-traffic-flows-per-day (Fig. 1a) and the flows-per-five-minutes of the day (Fig. 1b) for five days of consecutive Friday

Traffic flows

Define Qa,t as the traffic flow (i.e. the total number of vehicles) passing through the highway network within the area a during the time duration t. The area a can be a place with various spatial scales, ranging from a city, province or country, to any area with a certain specified size. Similarly, the duration t covers various temporal resolutions, e.g. from a few minutes or hours to an entire day. Let Qi,j as the traffic flow of each road link (link i) at the time j·Δ, where Δ is the updated

Estimating the effects based on traffic flows

After an event is detected, the impact of the disruption on traffic flows can be estimated based on the changes of this variable. Define Qa,t as the change ratio of Qa,t within the area a during the duration t; and it is calculated as follows.Qa,t=Qa,t-μω,Q,a,t-3σQ,a,t2+τω,Q,a,t2μω,Q,a,t-3σQ,a,t2+τω,Q,a,t2,ifQa,t<μω,Q,a,t-3σQ,a,t2+τω,Q,a,t2Qa,t-μω,Q,a,t+3σQ,a,t2+τω,Q,a,t2μω,Q,a,t+3σQ,a,t2+τω,Q,a,t2,ifQa,t>μω,Q,a,t+3σQ,a,t2+τω,Q,a,t20,else

The variable Qa,t characterizes the effect of the

The case study on the highway network of China

In this section, along with Sections 6–8, two real-world events are adopted to demonstrate the potential and effectiveness of the proposed method.

Incident identification

To examine the ability of the method in recognizing incidents, we choose the entire highway network as the study area, and utilize the traffic flow Qa,t each day or each hour, i.e. a=theentirecountry and t=eachdayoreachhour.

Incident impact analysis

The impact of the events is analyzed by means of the change ratios Qa,t and Ia,t of the traffic flows and congestion indexes, respectively. This analysis is conducted on three spatial levels, i.e. a=theentirecountry,eachprovince,oreachsinglelink, and in two temporal scales, i.e. t=eachdayoreachhour.

Sensitivity of the method on GPS data coverage

The sensitivity of the method in identifying incidents depends on a variety of factors, such as the extent of the disruption, quality of data processing, and spatial–temporal coverage of GPS data. The larger the disruptive extent, the higher the data processing quality and GPS data coverage, the more sensitive the detection is. In this study, given the level of data processing quality (e.g. 98.92% accuracy for map matching) and the inherent disruptive characteristics of the incidents (including

An additional application

In cooperation with the Highway Monitoring & Response Center in China (i.e. HMRC), we have applied the proposed method to the analysis of various other local or national incidents, including disasters (e.g. the Sichuan earthquake in 2019), social events (e.g. the Qingdao summit in 2018), major public health occurrences (e.g. the COVID-19 in 2020), and public holidays (e.g. the Labor Day in 2021). Different incidents share similarities but also have distinguished features, leading to varied

Conclusions and discussions

This paper has developed a spatial–temporal sensitive analysis method that accurately identifies an incident and estimates its disruptive effects on highway networks. In the method, two important indicators, including traffic flows and congestion indexes, along with their change ratios, are built to detect the incidents and evaluate the extent of the disruption.

There has been a number of traffic indicators (e.g. speeds, travel times, delays, congestion states, and levels of services) (Afrin and

CRediT authorship contribution statement

Zhenzhen Yang: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Writing - original draft, Writing - review & editing. Feng Liu: Methodology, Validation, Formal analysis, Writing - original draft, Writing - review & editing. Ziyou Gao: Supervision, Funding acquisition, Methodology, Writing - review & editing. Huijun Sun: Methodology, Writing - review & editing. Jiandong Zhao: Methodology, Writing - review & editing. Davy Janssens: Supervision, Project

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.

Acknowledgments

This work was supported by the Science Fund for Creative Research Groups of the National Natural Science Foundation of China [Grant number 71621001]; National Key Research and Development Program of China [Grant number 2018YFB2101003]; and the National Natural Science Foundation of China [Grant number 72101022, 71871011 and 72091513].

References (36)

Cited by (0)

View full text