Research Article  Open Access
Wu Ni, Wenbo Li, Hailei Wang, Chengxin Xiong, Dan Guo, "CapacityConstrained Contraflow Adaption for Lane Reconfiguration in Evacuation Planning", Journal of Advanced Transportation, vol. 2018, Article ID 2029586, 14 pages, 2018. https://doi.org/10.1155/2018/2029586
CapacityConstrained Contraflow Adaption for Lane Reconfiguration in Evacuation Planning
Abstract
This paper presents a heuristic contraflowbased reconfiguration evacuation algorithm, which is named CapacityConstrained Contraflow Adaption (CCAdap). First, it effectively calculates optimal candidate routes for evacuation. Second, an evaluation method is proposed for estimating these candidate routes. Third, CCAdap utilizes a contraflowbased method to reconfigure the evacuation routes to improve capacity constraints. Fourth, traffic conditions are updated in real time. Fifth, CCAdap reuses historical evacuation routes to reduce the computational cost and accelerate the evacuation process. Experimental results show that CCAdap generates highperforming evacuation strategies and can be used to tackle largescale evacuation planning.
1. Introduction
In recent years, both natural and manmade disasters have posed serious threats to humans, such as Hurricane Andrew [1], Hurricane Katrina [2], the “9/11” terrorist attack event [3], and 2011 Tohoku Earthquake Tsunami [4]. Effective evacuation plans are necessary for saving lives and minimizing casualties. The most important objective of an effective evacuation plan is to transfer evacuees to safe areas as quickly as possible. However, dynamic factors make this objective complicated to achieve. For example, the original intention in designing a transportation network is not to address the sharp increase of traffic flow when an emergency occurs [5]. The number of evacuees might far exceed the road capacity, thereby making evacuation route planning a computational challenge.
A wellknown routing algorithm, namely, CapacityConstrained Route Planner (CCRP) [1], was proposed for tackling evacuation route planning. CCRP models the capacity constraints of roads and intersections and selects the shortest paths for evacuation. Later, two heuristics, namely, Intelligent Load Reduction (ILR) and Incremental Data Structure (IDS), were introduced to improve the algorithm’s performance [6]. Although routing algorithms can effectively solve midsize traffic evacuation problems [6], they might not be effective for largescale evacuation scenarios with massive number of evacuees, which might lead to heavy traffic congestion, owing to capacity constraints. Contraflowbased methods are considered effective techniques for alleviating traffic congestion [2]. Contraflowbased methods reconfigure the edges in the ideal direction and reallocate edge capacity to reduce the evacuation time. For example, Kim et al. presented a heuristic approach that combines a Greedy method and a contraflowbased method for implementing transportation network reconfiguration and, to some extent, minimizing the total evacuation time [2]. However, finding an optimal contraflow plan for reconfiguring network is computationally challenging because we must enumerate all possible combinations of edges in the exhaustive search space [2]. Moreover, it takes a considerable amount of time to evaluate these contraflow candidates by calculating the evacuation time.
In this paper, we propose a CapacityConstrained Contraflow Adaption (CCAdap) algorithm for effective evacuation. CCAdap consists of five steps. First, we utilize an evacuation algorithm to generate optimal candidate routes for evacuees, which finds the routes with maximal flow rate among all the available routes from source vertices to sink vertices. Second, we propose an evaluation method for heuristically evaluating the candidate routes’ capabilities based on the traffic conditions and determining the appropriate routes for evacuation. Then, we implement a contraflowbased method for optimizing the routes’ performances and reducing traffic congestion. Next, traffic conditions are updated in real time to make the route planning more practical. Finally, CCAdap reuses the historical evacuation routes before calculating the available routes at each new time step. Experimental results indicate that CCAdap boosts performance in evacuation route planning.
The remainder of this paper is organized as follows: Section 2 introduces some related works. Section 3 defines the transportation network and contraflow problem. Section 4 details the CCAdap evacuation algorithm. Experimental results and analysis are presented in Section 5. Section 6 presents the conclusions of this paper.
2. Related Work
To the best of our knowledge, the existing methods can be divided into descriptive and prescriptive approaches [5]. Descriptive approaches aim at mimicking real emergency evacuations based on the simulation of the traffic situations and drivers’ behaviours, while prescriptive methods mainly focus on providing a highperforming evacuation plan with minimum evacuation time.
2.1. Descriptive Methods
The aim of descriptive methods is to simulate traffic evacuation situations as vividly as possible. Many traffic evacuation simulation tools have been proven to be capable of solving complex traffic evacuation problems, including MITSIMLab [7], VISSIM [8], MATSim [9], DynusT and DYNASMART [10], and AIMSUN [11]. In addition, some simulation methods utilize multimodel integration methods to simulate evacuees’ behaviours during evacuation and combine various existing evacuation methods for more realistic traffic simulation. For example, Pan et al. implemented a simulation framework for classifying evacuees’ behaviours into locomotion, steering, and social for evacuation analysis [12]. Wu and Huang combined a control volume model and a flow merging hypothesis to simulate evacuees’ behaviours in highrise building evacuation situations [13]. Noh et al. considered different behaviours of heterogeneous evacuees in building evacuation [14]. They presented a partially dedicated evacuation strategy that divides evacuees into heterogeneous groups, assigns corresponding routes based on a flow model and a simulation optimization approach, and, to some extent, minimizes the average evacuation time [14]. Beloglazov et al. studied the behaviours of people in wildfire evacuation and integrated traffic simulator, wildfire simulator, and behaviour simulator to define the risk metric and provided detailed plan of how evacuation unfold [15]. Yuan et al. constructed a multilevel agent decision model for simulating driver’s behaviours and determining actions for each agent [16]. These descriptive methods can achieve satisfactory performance in traffic evacuation for local communities. However, most exhibit poor generalization performance owing to high computational complexity.
2.2. Prescriptive Methods
Prescriptive methods typically provide suggested schedules for evacuation planners, aiming at reducing traffic congestion and minimizing total evacuation time. Zeng and Wang made a small modification to CCRP to improve the performance, which gives priority to longer evacuation routes for evacuating evacuees [17]. Kang et al. calculated the Dijkstra shortest evacuation paths and chose the best one for evacuees based on current traffic conditions [18]. Shahabi and Wilson proposed a CapacityAware Shortest Path Evacuation Routing (CASPER) framework for emergency traffic evacuation, which utilizes an advanced traffic model and an intelligent routing algorithm to calculate the optimal evacuation plan for evacuees [5]. Khan et al. used the Intelligent Transportation System (ITS) to compute the maximum traffic flow and the routes with least traffic congestion towards safe places for evacuees based on realtime traffic conditions [19]. Pourrahmani et al. considered the uncertain evacuee demand at the pickup source points by utilizing fuzzy credibility theory to present a genetic algorithm (GA) for handling the stated evacuation problem [20]. Chen et al. implemented a distribution Loadbalancing Emergency Guiding System (LEGS) to assign the fastest paths to transfer evacuees to the exits according to the capacity constraints and concurrent movement of evacuees [21]. Ikeda and Inoue proposed a MultiObjective Genetic Algorithm (MOGA) to evaluate evacuation routes by considering route distance and evacuation time to select optimal evacuation routes [22]. Heydar et al. divided the transportation network into pedestrian network and vehicular network, where pedestrians were assigned the shortest routes to the safe areas in pedestrian network and vehicles transferred evacuees to designated shelters in vehicular network [23]. These prescriptive methods have demonstrated the feasibility of calculating evacuation strategies for traffic evacuation. However, the performances of these prescriptive methods are usually limited owing to the capacity constraints of road segments and intersections.
Previous works [24–27] have considered the benefits of contraflowbased methods on evacuation planning. In addition, in the literatures, the optimality of evacuation planning by integrating contraflowbased methods has been validated. For example, Xie et al. designed a bilevel framework for optimizing network evacuation performance, in which the lower level solves the traffic assignment problem and the upper level integrates a contraflow method with crossing elimination strategies to optimize the traffic assignment [28]. Wang et al. implemented a multiobjective optimization model to determine the evacuation priorities and setup time for doing the contraflow operations [3]. Even et al. implemented a conflictbased pathgeneration algorithm to simultaneously guide the evacuation and select the tobereversed roads [29]. Kim et al. presented a Bottleneck Relief (BR) approach that iteratively flips mincut edges based on the theorem in [30] for solving massive traffic evacuation problems [2].
Abovementioned contraflowbased methods are selected to generate contraflow strategy that can tackle capacity constraints effectively [2, 3, 24–29]. However, there are two difficulties in calculating the appropriate contraflow strategy for the complex traffic system. The first difficulty is that it is a timeconsuming task to determine the appropriate tobereversed edges in the complex traffic network [2, 24, 25]. The second difficulty is that it is difficult to determine the ideal direction of these edges after being reversed [2, 3, 28, 29]. Thus, it is vital for the planners to find an appropriate contraflow strategy that could overcome these two difficulties.
So, this paper presents a prescriptive algorithm to generate appropriate contraflow strategy and optimal evacuation plan. For the first difficulty, the presented algorithm utilizes Greedy method and iterative optimization technique to select the tobereversed edges on the evacuation routes. For the second difficulty, the ideal direction of these tobereversed edges is determined by the traffic flow on the evacuation routes.
3. Modeling and Problem Formulation
Various methods, such as simulation [13–15], classification network [23], and mathematical modeling [1, 2], are used to formulate the evacuation situations. In this paper, the mathematical graph is used to describe the traffic evacuation situations [1, 2, 5, 6, 17].
3.1. Network Modeling
Suppose we are given a multisource and multisink transportation network , where and , respectively, represent the sets of vertices and edges. Here, is the source vertex set, is the transition vertex set, and is the sink vertex set. The definitions of the vertices and edges are given as follows:(i)Each vertex has a maximal vertex capacity and each source vertex has an extra initial occupancy , which represents the number of evacuees. For each sink vertex , the value of can be set according to the actual demand. In this paper, of each sink vertex is set to infinity.(ii)Each edge has a maximal edge capacity , a travel time , and an initial direction. is the number of evacuees (e.g., residents or vehicles) that can pass per unit period. The edge direction is initialized according to the real road conditions. is the time cost when evacuees pass edge . Various factors should be considered when formulating edge travel time, such as road length, traffic flow, and drivers’ behaviours. Many works have studied the relationship between these factors and edge travel time [13–15, 17]. This paper focuses on the impact of road length and traffic flow on the edge travel time. Considering the emergency situations, the traffic flow on the edges should be limited by the maximal edge capacity. Thus, the evacuees can use maximal speed passing one given edge when the traffic flow does not exceed the maximal edge capacity. In other words, more evacuees can be evacuated to safe area using this strategy in the minimum time. According to this, the edge travel time in the transportation network model is determined by the length and the maximal edge capacity of one given road.
There are different traffic elements in transportation network including buildings, crossroads, playgrounds, shelters, and parks. Buildings, crossroads, and playgrounds are the most common traffic elements in transportation network. Figure 1 illustrates the modeling results of the buildings, the crossroad, and the playground. Firstly, the buildings are modeled as source vertex. Secondly, the crossroad is modeled as transition vertex. Thirdly, the playground is modeled as sink vertex. The edges of the vertices are modeled according to the traffic configuration. As for the other traffic elements, they are modeled similar to these three common traffic elements shown in Figure 1.
(a) Source vertex
(b) Transition vertex
(c) Sink vertex
The objective of transportation evacuation is to find an optimal plan with minimum evacuation time. The objective function is defined in
is the total evacuation time, is the initial time step when a group of evacuees leaves the source vertex, is the tobesolved evacuation route set, and is an evacuation route in .
is used to calculate the travel time of route when evacuees pass through the edges along route and is expressed in (2). For route , where is the th vertex of route , is the source vertex of route , and is the sink vertex of route . means the travel time from vertex to vertex along route .
is used to calculate the maximal traffic flow of route under capacity constraints of vertices and edges and is expressed in (3). denotes the available residual capacities of vertex on route at its arrival time , where , and denotes the available residual capacity of edge on route at its arrival time .
Formula (1) illustrates the capacity constraints along the time dimension during evacuation. Constraint ensures that the outflow of source vertex does not exceed the initial occupancy of source vertex . Constraint enforces that the inflow of the sink vertices must be equal to the initial occupancy of the source vertices. In addition, the model capacity of vertices and edges cannot be negative during evacuation due to Constraint .
3.2. Contraflow Problem
In a large transportation network, long distance traffic jams usually cover the roads when emergencies occur. Contraflowbased methods are considered effective techniques for remedying traffic congestion [2]. However, an appropriate contraflow reconfiguration is difficult to calculate. Generally, two main difficulties are encountered in calculating optimal contraflow reconfiguration: (1) it is NPhard to find all possible combinations of edge directions and calculate the evacuation time of these contraflow combinations and (2) it is timeconsuming to evaluate the effects of dynamic traffic flow on these contraflow reconfiguration candidates.
Figure 2 illustrates an evacuation scenario and two contraflowbased evacuation planning results. The evacuation time is 8 time units based on the original network configuration in Figure 2(a). Figures 2(b) and 2(c) illustrate two contraflow reconfigurations, which combine all the twoway edge directions and merge the edge capacities to increase the upper bound on the evacuation capacity. The contraflow reconfiguration in Figure 2(b) reduces the evacuation time to 5 time units, while the contraflow reconfiguration in Figure 2(c) reduces the evacuation time to 7 time units. The only difference between the two contraflow reconfigurations is the edge direction between vertex and vertex . Thus, it is vital for an optimal contraflow reconfiguration to select the critical edges that affect evacuation performance and determine their ideal direction after reversal.
(a) Original network
(b) Evacuation planning 1
(c) Evacuation planning 2
4. Proposed Algorithm
This paper proposes the CCAdap algorithm, which heuristically evaluates the evacuation routes, and integrates it with a contraflowbased method for distributing the evacuation routes with the least traffic congestion and maximal flow rate to evacuees. In addition, CCAdap reuses historical evacuation routes to reduce computational costs and accelerate the evacuation process. Algorithm 1 gives the pseudocode of the CCAdap algorithm.

Step 0 (initialization). For a given network , let be the tobereversed edge set, be the initial time step when evacuees leave source vertices, be the tobesolved evacuation planning route set, and be the historical evacuation route set.
Step 1 (GetEvaRoute). CCAdap utilizes the evacuation route generation method in the MaxFlow Rate Priority (MFRP) algorithm [31] to select optimal candidate route from among the available routes from source vertices to the nearest sink vertices. MFRP introduces parameter to describe the route capacity. The formula of is shown in (4).The main strategy of the MFRP algorithm [31] is to calculate the available candidate routes by Dijkstra algorithm and balance the number of transferred evacuees with the travel time to select evacuation routes with maximal . The performance of MFRP has been analyzed in [31]. CCAdap guarantees the shortest possible evacuation time by employing MFRP [31] to select the optimal candidate routes.
Step 2 (RouteEvaluation). The use of lowquality routes might aggravate traffic congestion and increase the total evacuation time. CCAdap introduces a parameter for quantizing route , which is generated in Step 1, and historical evacuation route to avoid selecting lowerquality evacuation routes. The formula of is presented inwhere is the historical evacuation route in , which has the same source vertex as . Equation (5) represents the evacuation capability gap between routes and when they evacuate evacuees at same times under the same traffic conditions.
The RouteEvaluation function operates as follows: First, we introduce parameter for reserving the evaluation result and set initially. For each historical evacuation route in , we evaluate between and by employing (5), where and have the same source vertex, and reserve the evaluation result using as follows, which is expressed in (6): If is less than zero, then we set and calculate the next historical evacuation route in . Otherwise, is suitable for evacuation and we set and terminate the evaluation procedure. If there is no route with the same source vertex as in , we assume that can also be used for evacuation and set .
Step 3 (ContraflowReconfiguration). Greedy method can be used in evacuation route planning [2]. The appropriate contraflow strategy can be determined using Greedy method which has the advantage of obtaining the global optimal solution by local optimal solution. Thus, CCAdap integrates Greedy method and iterative optimization technique to reverse the twoway edges on the evacuation routes iteratively in this paper. However, the edges on the evacuation routes cannot be reversed repeatedly as long as the edges have been flipped. The modified edge direction of the tobereversed edges depends on the traffic flow along the evacuation routes. The upper bound on the capacity of the flipped edges is increased by merging the twoway edges. The travel time of the reversed edges is equal to that of the twoway edges, which have the same edge direction as the flipped edges.
Figure 3 illustrates the steps for implementing the proposed contraflow reconfiguration method. Figure 3(a) is the original network configuration. In Figure 3(b), we select evacuation route . Its traffic flow is . As shown in Figure 3(c), the twoway edges on this route are merged and the maximal capacity of the flipped edges is increased. After contraflow reconfiguration, the traffic flow is increased to . Thus, the evacuation capability of route is improved.
(a) Original network
(b) Select evacuation route
(c) Contraflow result
Step 4 (NetworkUpdate). After evacuees are transferred from source vertex to sink vertex, the available residual capacities of transition vertices and edges along route are updated in (7) and (8), respectively, and the source vertex’s residual evacuees of route are updated in (9).Based on (7), (8), and (9), the traffic conditions can be monitored in real time for the nextiteration route calculation.
Step 5 (HRouteReuse). When there is no available evacuation route at the current time step, CCAdap proceeds to the next time step (++) and returns to Step 1 to calculate routes. Before calculating routes, CCAdap reuses historical evacuation routes in to reduce the computational cost and accelerate the evacuation process. To avoid evacuating along the same evacuation routes continually, CCAdap introduces a parameter that marks historical evacuation route in if is reserved at the last time step. is defined inThe HRouteReuse function operates as follows: For each historical evacuation route in , if under the current traffic conditions, it will be excluded from . Otherwise, if the corresponding is equal to 0, is reserved at the last time step and is not suitable for evacuation at the current time step, we set for the next iteration. If is equal to 1, then is used for evacuation at the current time step and traffic conditions are updated by Step 4.
In each iteration, the generated route in Step 1 is reserved in and when is used to transfer evacuees and of is set to 0. Meanwhile, the tobereversed edges on route are reserved in . To complete the evacuation, Steps 1–5 can be repeated until there is no evacuee who needs to be evacuated.
5. Experiment
5.1. Experimental Design
There are many evacuation algorithms that can provide evacuation route planning, such as the classical CCRP algorithm [1], LRP algorithm [17], MR algorithm [31], MFRP algorithm [31], and Greedy algorithm [2]. Reference [31] has validated that the performance of MFRP algorithm is better than LRP and MR. Besides, Greedy algorithm is a classical contraflow evacuation algorithm. CCAdap algorithm integrates MFRP algorithm and Greedy algorithm in this paper. Thus, CCRP, MFRP, and Greedy algorithms are selected to validate the performance of CCAdap algorithm.
5.2. Experiment Data
In this paper, we use the following notations to describe evacuation situations: is the number of vertices, is the number of source and sink vertices, is the number of edges, and is the number of evacuees. The experimental data that are used in this paper is generated by the network simulator tool NETGEN [32]. However, NETGEN does not generate the maximal vertex capacity for transition vertices. This paper uses a random number generation function, namely, rand(), to generate the maximal vertex capacity for transition vertices in (11). The seed of the rand() function is set to the vertex number to avoid generating the same networks. and are two coefficients in (11), which imply that the lower bound on the edge capacity is and the upper bound on the edge capacity is . and can be reset according to the practical requirements. In this paper, we set and [31].
5.3. Experimental Result
5.3.1. Feasibility Verification
The purpose of this section is to evaluate the feasibility of using CCAdap in evacuation planning. CCAdap is compared with CCRP and MFRP to determine whether CCAdap can be used for evacuation planning and to optimize evacuation plans by applying a contraflowbased method.
In this section, two experiments are carried out. Each experiment has five test groups and the number of vertices, edges, source, and sink vertices is fixed. We vary the number of evacuees from 20000 to 100000. Figures 4 and 5 depict the results on the evacuation time and run time of the three algorithms in smallscale and largescale transportation networks, respectively. According to Figure 4(a), CCAdap has the smallest and nearly linearly growing evacuation time with the increase of the number of evacuees, while CCRP and MFRP have steeply increasing evacuation time and the evacuation time of CCRP is almost double that of CCAdap. The run time of CCRP and MFRP grows rapidly as the number of evacuees increases, while CCAdap has a stable run time and the run time of CCRP and MFRP is more than twice that of CCAdap when the number of evacuees is 100000 in Figure 4(b). As shown in Figure 5, CCAdap has the shortest evacuation time with the increase in the number of evacuees in Figure 5(a), followed by CCRP, and MFRP has the longest evacuation time. In terms of the run time, CCAdap, CCRP, and MFRP all show nearly linear growth with the increase in the number of evacuees, and CCAdap has a run time that is nearly half those of CCRP and MFRP.
(a) ; ;
(b) ; ;
(a) ; ;
(b) ; ;
5.3.2. Scalability Verification
The purpose of this section is to evaluate the scalability of CCAdap to complex evacuation situations and CCAdap is compared with CCRP, MFRP, and Greedy algorithms.
Comparison on Networks with Different Number of Edges. Figure 6 plots the experimental results in terms of evacuation time and run time of the four algorithms against the number of edges. This experiment is carried out with five subexperiments. The number of vertices is fixed at 256, and the number of source and sink vertices is fixed at 26. The number of evacuees is varied from 20000 to 100000. Each subexperiment has two test groups and the number of edges is varied from 768 to 1528. According to Figure 6, CCAdap has the shortest evacuation time in each subexperiment as the number of edges varies. In addition, CCAdap shows the minimum change in evacuation time when the number of edges varies, followed by Greedy, MFRP, and CCRP. Regarding the run time of these four algorithms when the number of edges varies, as shown in Figure 6, CCAdap has the shortest run time comparing with the CCRP, MFRP, and Greedy algorithms; the run time of CCAdap is less than 20 percent of those of CCRP and Greedy. In addition, the number of edges has little effect on the run time of CCAdap, while the CCRP, MFRP, and Greedy algorithms are affected by the variation of the number of edges.
(a) ; ;
(b) ; ;
(c) ; ;
(d) ; ;
(e) ; ;
(f) ; ;
(g) ; ;
(h) ; ;
(i) ; ;
(j) ; ;
Comparison on Different Number of Evacuees. Figure 7 plots the experimental results on evacuation time and run time of the four algorithms against the number of evacuees. This experiment is carried out with three subexperiments, and the numbers of vertices, edges, and source and sink vertices are fixed. Each subexperiment has five test groups and the number of evacuees is varied from 20000 to 100000. According to Figure 7, the evacuation time of CCAdap increases (perhaps linearly) with the increase in the number of evacuees, while CCRP has a steeply increasing evacuation time, MFRP and Greedy show volatility growth of evacuation time with the increase in the number of evacuees. In particular, according to Figures 7(c) and 7(e), the evacuation time of CCAdap is almost 50 percent those of CCRP, MFRP, and Greedy when tackling a massive number of evacuees. Moreover, CCAdap has a short and stable run time with the increase in the number of evacuees, while the CCRP, MFRP, and Greedy algorithms have steeply, perhaps superlinearly, increasing run time. The run time of CCAdap is almost 10 percent that of the Greedy algorithm in Figures 7(b), 7(d), and 7(f) and that of CCRP in Figures 7(b) and 7(d) when tackling a massive number of evacuees.
(a) ; ;
(b) ; ;
(c) ; ;
(d) ; ;
(e) ; ;
(f) ; ;
Comparison on Different Numbers of Source and Sink Vertices. Figure 8 plots the experimental results on the evacuation time and run time of the four algorithms against the number of source and sink vertices. This experiment is carried out with five subexperiments. The number of vertices is fixed at 256, the number of edges is fixed at 1528, and the number of evacuees is varied from 20000 to 100000. Each subexperiment has two test groups and the number of source and sink vertices is varied from 26 to 52. According to Figure 8, CCAdap produces the solutions with shortest evacuation time comparing with the CCRP, MFRP, and Greedy algorithms when the number of source and sink vertices is varied. In addition, the evacuation time of CCAdap has the slowest increase with the increase in the number of source and sink vertices comparing with the other three algorithms. For different number of source and sink vertices, CCAdap has the fastest run time comparing with the CCRP, MFRP, and Greedy algorithms. As the number of source and sink vertices varies from 26 to 52, the run time of the MFRP and Greedy algorithms rises steeply, CCRP shows a small growth in run time, and CCAdap shows the smallest increase in run time.
(a) ; ;
(b) ; ;
(c) ; ;
(d) ; ;
(e) ; ;
(f) ; ;
(g) ; ;
(h) ; ;
(i) ; ;
(j) ; ;
Comparison on Large Transportation Networks. Figure 9 depicts the experimental results of the four algorithms in terms of evacuation time and run time on a largescale transportation network. The experiment is carried out with five test groups and has a fixed network configuration of 1024 vertices, 5258 edges, and 102 source and sink vertices. We vary the number of evacuees from 20000 to 100000 to test the algorithms’ performances. As shown in Figure 9(a), CCAdap can generate highquality evacuation solutions with the shortest evacuation time when the number of evacuees grows in a largescale transportation network. Moreover, the run time of CCAdap shown in Figure 9(b) also has the smallest and most stable growth with different number of evacuees in largescale transportation network evacuation scenarios. In addition, CCAdap has the minimum growth ratio of run time with the increase in the number of evacuees, comparing with CCRP, MFRP, and Greedy (which have growth ratios of nearly 50 percent).
(a) ; ;
(b) ; ;
5.4. Summary of Experiments
There are many dynamic factors that are critical to the transportation network, such as number of evacuees and affected areas, which makes evacuation management a daunting task. These dynamic factors can be evaluated systematically by building appropriate evacuation models [33]. Only by taking these factors into consideration we can optimize evacuation management. This section will discuss these dynamic factors and evaluate the optimality of the CCAdap algorithm based on the experimental results.(i)Number of edges: the more edges there are, the more difficult it is to find appropriate routes to evacuate evacuees. CCAdap can intelligently evaluate the routes to choose the optimal one for evacuation, thereby making it more scalable with the number of edges.(ii)Number of source and sink vertices: an increase in the number of source and sink vertices in a transportation network means that the evacuees are more scattered in the network [1]. It might lead to difficulty in calculating the evacuation routes. CCAdap reuses the historical evacuation routes to reduce the computational costs, thereby making CCAdap more scalable with the number of source and sink vertices.(iii)Number of evacuees: capacity constraints are the major limiting factor in optimizing an evacuation strategy. Traffic congestion occurs easily when there are too many evacuees. CCAdap applies a contraflowbased method to improve the performance of evacuation routes, which makes CCAdap more scalable with the number of evacuees.(iv)Traffic scale: large and complex urban evacuation problems cause most evacuation route planning approaches to fail. CCAdap integrates an iterative optimization technique and a contraflowbased method to optimize an evacuation plan, thereby making CCAdap more effective in addressing urban evacuation planning.
6. Conclusions
This paper presents a heuristic contraflow reconfiguration evacuation algorithm, which is called CapacityConstrained Contraflow Adaption algorithm. CCAdap can heuristically evaluate evacuation routes to assign highquality routes to evacuees. Moreover, the proposed contraflowbased method is applied to reconfigure evacuation routes to reduce traffic congestion and improve evacuation performance. Meanwhile, CCAdap reuses historical evacuation routes to reduce computational costs and accelerate the evacuation process. Experimental results have shown that the CCAdap algorithm can optimize evacuation route planning in terms of evacuation time and run time.
However, more research is required for CCAdap. First, a more effective evacuation route generation method should be constructed. Second, the evaluation method should be improved to evaluate the evacuation routes more effectively.
Data Availability
The data used to support the findings of this study are available from the first author.
Conflicts of Interest
The authors declare that they have no conflicts of interest.
Acknowledgments
This work is supported by the National Key Research and Development Program of China (no. 2016YFB0502600) and the National Natural Science Foundation of China (no. 41701594).
References
 Q. Lu, B. George, and S. Shekhar, “Capacity constrained routing algorithms for evacuation planning: A summary of results,” in Proceedings of the International Symposium on Spatial and Temporal Databases, pp. 291–307, Berlin, Germany, 2005. View at: Google Scholar
 S. Kim, S. Shekhar, and M. Min, “Contraflow transportation network reconfiguration for evacuation route planning,” IEEE Transactions on Knowledge and Data Engineering, vol. 20, no. 8, pp. 1115–1129, 2008. View at: Publisher Site  Google Scholar
 J. W. Wang, H. F. Wang, W. J. Zhang, W. H. Ip, and K. Furuta, “Evacuation planning based on the contraflow technique with consideration of evacuation priorities and traffic setup time,” IEEE Transactions on Intelligent Transportation Systems, vol. 14, no. 1, pp. 480–485, 2013. View at: Publisher Site  Google Scholar
 T. Takabatake, T. Shibayama, M. Esteban, H. Ishii, and G. Hamano, “Simulated tsunami evacuation behavior of local residents and visitors in Kamakura, Japan,” International Journal of Disaster Risk Reduction, vol. 23, pp. 1–14, 2017. View at: Publisher Site  Google Scholar
 K. Shahabi and J. P. Wilson, “CASPER: Intelligent capacityaware evacuation routing,” Computers, Environment and Urban Systems, vol. 46, pp. 12–24, 2014. View at: Publisher Site  Google Scholar
 S. Kim, B. George, and S. Shekhar, “Evacuation route planning: scalable heuristics,” in Proceedings of the 15th annual ACM international symposium on Advances in geographic information systems, p. 20, 2007. View at: Google Scholar
 M. Jha, K. Moore, and B. Pashaie, “Emergency evacuation planning with microscopic traffic simulation,” Transportation Research Record, no. 1886, pp. 40–48, 2004. View at: Google Scholar
 M. Fellendorf and P. Vortisch, “Microscopic traffic flow simulator VISSIM,” in Fundamentals of traffic simulation, pp. 63–93, Springer, New York, NY, USA, 2010. View at: Publisher Site  Google Scholar
 G. Lämmel, D. Grether, and K. Nagel, “The representation and implementation of timedependent inundation in largescale microscopic evacuation simulations,” Transportation Research Part C: Emerging Technologies, vol. 18, no. 1, pp. 84–98, 2010. View at: Publisher Site  Google Scholar
 Y.T. Hsu and S. Peeta, “Clearance Time Estimation for Incorporating Evacuation Risk in Routing Strategies for Evacuation Operations,” Networks and Spatial Economics, vol. 15, no. 3, pp. 743–764, 2015. View at: Publisher Site  Google Scholar
 V. Zyryanov and A. Feofilova, “Simulation of Evacuation Route Choice,” Transportation Research Procedia, vol. 20, pp. 740–745, 2017. View at: Google Scholar
 X. Pan, C. S. Han, K. Dauber, and H. H. Law, “A multiagent based framework for the simulation of human and social behaviors during emergency evacuations,” Ai & Society, vol. 22, no. 2, pp. 113–132, 2007. View at: Google Scholar
 G.Y. Wu and H.Ch. Huang, “Modeling the emergency evacuation of the high rise building based on the control volume model,” Safety Science, vol. 73, no. 5, pp. 62–72, 2015. View at: Publisher Site  Google Scholar
 D.J. Noh, J. Koo, and B.I. Kim, “An efficient partially dedicated strategy for evacuation of a heterogeneous population,” Simulation Modelling Practice and Theory, vol. 62, pp. 157–165, 2016. View at: Publisher Site  Google Scholar
 A. Beloglazov, M. Almashor, E. Abebe, J. Richter, and K. C. B. Steer, “Simulation of wildfire evacuation with dynamic factors and model composition,” Simulation Modelling Practice and Theory, vol. 60, pp. 144–159, 2016. View at: Publisher Site  Google Scholar
 S. Yuan, S. A. Chun, B. Spinelli, Y. Liu, H. Zhang, and N. R. Adam, “Traffic evacuation simulation based on multilevel driving decision model,” Transportation Research Part C: Emerging Technologies, vol. 78, pp. 129–149, 2017. View at: Publisher Site  Google Scholar
 M. Zeng and C. Wang, “Evacuation route planning algorithm: Longer route preferential. International Symposium on Neural Networks,” in Proceedings of the International Symposium on Neural Networks, pp. 1062–1071, Berlin, Germany, 2009. View at: Google Scholar
 W. Kang, F. Zhu, Y. Lv, G. Xiong, L. Xie, and B. Xi, “A heuristic implementation of emergency traffic evacuation in urban areas,” in Proceedings of the 2013 IEEE International Conference on Service Operations and Logistics, and Informatics, SOLI 2013, pp. 40–44, chn, July 2013. View at: Google Scholar
 M. U. S. Khan, O. Khalid, Y. Huang et al., “MacroServ: A Route Recommendation Service for LargeScale Evacuations,” IEEE Transactions on Services Computing, vol. 10, no. 4, pp. 589–602, 2017. View at: Publisher Site  Google Scholar
 E. Pourrahmani, M. R. Delavar, and M. A. Mostafavi, “Optimization of an evacuation plan with uncertain demands using fuzzy credibility theory and genetic algorithm,” International Journal of Disaster Risk Reduction, vol. 14, pp. 357–372, 2015. View at: Publisher Site  Google Scholar
 L.W. Chen, J.H. Cheng, and Y.C. Tseng, “Distributed Emergency Guiding with Evacuation Time Optimization Based on Wireless Sensor Networks,” IEEE Transactions on Parallel and Distributed Systems, vol. 27, no. 2, pp. 419–427, 2016. View at: Publisher Site  Google Scholar
 Y. Ikeda and M. Inoue, “An Evacuation Route Planning for Safety Route Guidance System after Natural Disaster Using Multiobjective Genetic Algorithm,” in Proceedings of the 20th International Conference on Knowledge Based and Intelligent Information and Engineering Systems, KES 2016, pp. 1323–1331, gbr, September 2016. View at: Google Scholar
 M. Heydar, J. Yu, Y. Liu, and M. E. H. Petering, “Strategic evacuation planning with pedestrian guidance and bus routing: a mixed integer programming model and heuristic solution,” Journal of Advanced Transportation, vol. 50, no. 7, pp. 1314–1335, 2016. View at: Publisher Site  Google Scholar
 U. Pyakurel, T. N. Dhamala, and T. N. Dhamala, “Models and algorithms on contraflow evacuation planning network problems,” International Journal of Operations Research, vol. 12, no. 2, pp. 36–46, 2015. View at: Google Scholar
 U. Pyakurel and T. N. Dhamala, “Continuous Dynamic Contraflow Approach for Evacuation Planning,” Annals of Operations Research, vol. 253, no. 1, pp. 573–598, 2017. View at: Publisher Site  Google Scholar
 U. Pyakurel and T. N. Dhamala, “Continuous Time Dynamic Contraflow Models and Algorithms,” Advances in Operations Research, vol. 2016, Article ID 7902460, 7 pages, 2016. View at: Publisher Site  Google Scholar
 J. Zhao, Y. Liu, and P. Li, “A network enhancement model with integrated lane reorganization and traffic control strategies,” Journal of Advanced Transportation, vol. 50, no. 6, pp. 1090–1110, 2016. View at: Publisher Site  Google Scholar
 C. Xie, D.Y. Lin, and S. Travis Waller, “A dynamic evacuation network optimization problem with lane reversal and crossing elimination strategies,” Transportation Research Part E: Logistics and Transportation Review, vol. 46, no. 3, pp. 295–316, 2010. View at: Publisher Site  Google Scholar
 C. Even, V. Pillac, and P. Van Hentenryck, “NICTA evacuation planner: Actionable evacuation plans with contraflows,” Frontiers in Artificial Intelligence and Applications, vol. 263, pp. 1143–1148, 2014. View at: Google Scholar
 D. R. Fulkerson, “Flows in networks,” in Recent advances in mathematical programming, pp. 319–331, McGrawHill, New York, NY, USA, 1963. View at: Google Scholar  MathSciNet
 D. Guo, C. Gao, W. Ni, and X. Hu, “MaxFlow rate priority algorithm for evacuation route planning,” in Proceedings of the 1st IEEE International Conference on Data Science in Cyberspace, DSC 2016, pp. 275–283, chn, June 2016. View at: Google Scholar
 D. Klingman, A. Napier, and J. Stutz, “NETGEN: a program for generating large scale capacitated assignment, transportation, and minimum cost flow network problems,” Management Science, vol. 20, no. 5, pp. 814–821, 1974. View at: Publisher Site  Google Scholar
 E. Ronchi, F. N. Uriz, X. Criel, and P. Reilly, “Modelling largescale evacuation of music festivals,” Case Studies in Fire Safety, vol. 5, pp. 11–19, 2016. View at: Publisher Site  Google Scholar
Copyright
Copyright © 2018 Wu Ni et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.