See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/257717210

A Method for Evacuation Route Planning in Disaster Situations

Article  in  Procedia - Social and Behavioral Sciences · October 2012

DOI: 10.1016/j.sbspro.2012.09.768

CITATIONS READS

50 1,184

3 authors, including:

Vania Campos Renata Albergaria de Mello Bandeira

Instituto Militar de Engenharia (IME) Instituto Militar de Engenharia (IME)
53 PUBLICATIONS   259 CITATIONS    55 PUBLICATIONS   269 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Transporte de Carga View project

URBAN MOBILITY View project

All content following this page was uploaded by Vania Campos on 07 June 2016.

The user has requested enhancement of the downloaded file.

 15th Edition of the European Working Group on Transportation – EWGT 2012

A method for evacuation route planning in disaster situations

Vania Camposa, Renata Bandeiraa, Adriano Bandeiraa

a
Instituto Militar de Engenharia, Praça General TIburcio 80, Praia Vermelha, 22290-270, Rio de Janeiro, Brasil

Abstract

In this paper, the design and analysis of evacuation routes in transportation networks in cases of natural disasters is
examined. A method for evacuation route planning in disaster situations is suggested. This method applies iteratively
a heuristic algorithm to define two independent paths from the disaster area to each shelter for vehicle flow allocation
in evacuation planning, considering both travelling time and capacity of the transportation network as parameter for
analysis. Routes to different shelters cannot present intersection points either in order to allow continuous traffic flow
and reduce potential accidents. An evacuation planning exercise is also presented.

Keywords: Disasters, Emergency planning, Evacuation planning, Routing.

1. Introduction

Factors such as global warming, environmental degradation and increasing urbanization expose a greater
number of people to the threat of natural disasters. In the last three decades, the rate of disasters has risen
from 50 to 400 per year (Kovacs and Spens, 2009) and it is still expected to increase five times more on
the next 50 years (Thomas and Kopczack, 2007). In 2010, 207 million people suffered from disasters,
which caused 296,800 deaths and losses of 109 billion dollars (Sapir, 2011).
These natural disasters require the immediate mobilization and action of multiple stakeholders due to
the unexpected nature and amplitude of the event, the diversity and quantity of supplies and services
needed by the victims (Chandes and Paché, 2009). The focus of these disaster relief operations is to
“design the transportation of first aid material, food, equipment, and rescue personnel from supply points
to a large number of destination nodes geographically scattered over the disaster region and the
evacuation and transfer of people affected by the disaster to the healthcare centers and shelters safely and
rapidly” (Barbarosoglu et al., 2004). In this context, logistics for emergency relief operations is known as
humanitarian logistics. Transporting the critical supplies and personnel from source to affected areas,
evacuation of all types of individuals to shelters and medical facilities wherever necessary play major
roles in humanitarian logistics (Apte, 2009).
Evacuation can be mandatory, recommended, or voluntary, besides it may differ by scale, objects of
relocation (people vs. property), and by level of control by authorities (Stepanov and Smith, 2009).
Depending on the disaster type, pre-warning of sudden-onset disasters can leave enough time for
evacuations prior to the event (Kovacs and Spens, 2007). Another influential factor in the case of natural
disasters is the probability of their occurrence and effects. Seasonal disasters, even if sudden, pose lesser
problems than non-seasonal disasters, even though evacuation plans can be in place for either situation
(Apte, 2009). Emergency evacuation plans assign evacuees to fixed routes or destinations even before the
disasters. An evacuation plan defines optimal evacuation policies for the population from areas under risk
and uncertainty (Stepanov and Smith, 2009). A lesson learned due to the evacuation fiasco after
2 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000

Hurricane Katrina in the United States is the need for evacuation planning (Apte, 2009), even though real-
time design or re-evaluation of these plans may be required after the disaster strikes (Alexander, 2002).
Techniques and models have been developed to simulate the evacuation of people living near the
affected area, with the main objective of identifying the problems that can happen in the moment of
evacuation. These problems can be vehicular congestion and accidents that contribute to increasing the
time of evacuation and the number of injuries. However, conventional methods and heuristics for defining
evacuation routes generally are based mainly on geographic proximity and seek for the shortest travelling
time (Han et al., 2006). Such techniques do not guarantee that the capacity of the routes will satisfy the
intense demand for transportation during evacuation, and neither that the resulting routes will not present
coincident intersection points, which could be eventual bottlenecks susceptible to potential accidents.
Therefore, this paper presents a method for defining two independent paths from the disaster area (origin)
to each destination points (shelters) for vehicle flow allocation in evacuation planning, considering both
travelling time and capacity of the transportation network as parameters for analysis. This type of routes
has no intersection points (they are called disjoint paths). Besides, routes to different destinations points
(shelters) cannot have intersection points as well, so their use minimizes the problem of accidents and
permits a continuous traffic flow.
The application of this method allows the identification of the best possible independent evacuation
routes from a potential disaster area to different shelters, previously defined. The method is
applied iteratively, allowing the removal of intersection points among the different routes. However, in
cases which a set of routes without intersection points is unfeasible, the method provides previous
knowledge of this intersection point, indicating the need for interventions at this point in order to avoid
potential accidents caused by conflicting movements.
The paper proceeds as follows. Section 2 provides a literature review on evacuation planning. Section
3 describes the method proposed. Sections 4 present the results of the application of the method. Finally,
Section 5 concludes and summarizes insights for both academia and logistics practitioners.

2. Evacuation planning

Evacuation is a complex process consisting of several consecutive phases. After the detection of an
incident, decision makers evaluate the potential threat for specific areas and then issue an evacuation
order for these areas if the risk is significant and there are no shelters to provide adequate in-place
protection (Stepanov and Smith, 2009). The type of disaster will dictate total or partial evacuation to
distant or near-by relief location and if it is long term or temporary process (Apte, 2009). Evacuation alert
is communicated to the population, who makes a decision to evacuate or not depending on their
perception of danger and then the population is transferred through a transportation network to designated
safe areas (Stepanov and Smith, 2009). Condition of the infrastructure will force which mode of transport
to use but the evacuation also depends on the available fleet of vehicles, their capacity and on
circumstances surrounding the evacuee (Apte, 2009). Finally, evacuees arrive to areas outside of danger
and a verification that they have made it safely must be carried out (Stepanov and Smith, 2009).
The evacuation time may range from hours to weeks or even months (Church and Sexton, 2002, apud
Stepanov and Smith, 2009) depending on the scale of the disaster. Han et al. (2006) suggests that the
major problem in evacuation operations is that evacuation-zone exit routes are often limited in number
and insufficient in capacity to handle the traffic surge during a large-scale emergency evacuation.
Stepanov and Smith (2009) reinforces that capacity of transportation networks generally cannot satisfy
the intense demand for transportation during an evacuation, and that even on small neighborhood-scale
evacuations, transportation networks can impede the fast clearing of the population from an effected area.
In order to manage such emergencies more effectively, decision makers may benefit from having in-
place evacuation plans for scenarios which are most likely to happen, even though real-time design or re-
evaluation of evacuation plans may be required after the disaster strikes (Alexander, 2002). Therefore,
 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000 3

studies have focused on methods to improve the planning and operational aspects of the evacuation
process to maximize the utility of the existing transportation network (Han et al., 2006). Apte (2009)
presents a summary of some of the research in evacuation. Stepanov and Smith (2009) also reviewed a
series of evacuation models and proposed that modelling techniques can be grouped by computation
techniques into analytical and simulation techniques. From a review of the evacuation models, Stepanov
and Smith (2009) conclude that two approaches are evident: (i) the first approach defines a set of optimal
routes and evaluates performance measures simultaneously; and (ii) a second approach uses an analytical
optimization technique to offer a routing policy, and then this policy is evaluated with traffic simulation
model. According to Stepanov and Smith (2009), the first approach is prevalent in practice.
In this context, this paper presents a heuristics algorithm that can be applied for defining independent
routes for developing in-place evacuation plans. According to Stepanov and Smith’s (2009) classification,
the algorithm proposed applies analytical techniques to define a set of optimal routes and evaluate
performance measures simultaneously. The characteristics of the method are presented on the following
section.

3. Characteristics of the Method

According to Graph theory (Boaventura, 1996; Bollobás, 1978), there is a limited number (k) of
independent paths (called disjoint paths) between a pair of nodes in a network; however it is possible to
exist several sets (of k elements) of these paths in the same network.
The developed algorithm allows identifying the best set of independent routes (disjointed paths
without intersection points) from the affected area (origin) to each destination point (pre-defined shelters),
considering both travelling time and capacity of the transportation network as parameters for analysis. For
each destination point (shelter), two independent paths will be identified. We chose to select two
independent paths from the origin to each destination point for different reasons. In the first place, it is
important to have an alternative route in cases of infrastructure failure or road blocks. The evacuation
plan can also designate different flows for each path, so there will be a path with the flow from the
destination point to the shelter to which evacuation can be allocated and another with the flow from the
shelter to the disaster area to which the transport of equipments, critical supplies and personnel from
source to affected areas can be allocated. A circular flow can also be established, especially in cases of
large-scale emergency evacuation, so empty vehicles can travel from the shelter to the disaster area in one
path and the other path will be used to the transportation of victims to the disaster areas. This measure can
help to increase the speed of the evacuation.
In the process of selecting these two independent paths, an index defined by the ratio C/T (capacity
/travel time in the route) is determined for each route, and the best route set is that which presents the
largest sum of these indexes. The reason to choose this index is because comparatively a path is better
than other if it has larger capacity or a smaller travel time. If two paths have the same capacity, the path
which has the smallest travel time should be chosen (its index is larger) and if they have the same time the
one with the largest capacity is considered better. In summary, considering the possibility of existing
independent k-paths between a pair of source and destination node in a network, the method determines a
set of routes with the objective of maximize the summation of the ratio of each path, that is:
k
Max  Cp / Tp
p 1
p = 1 ...k (1)

Besides, routes to different destinations points (shelters) cannot have intersection points as well, so
their use minimizes the problem of accidents and permits a continuous traffic flow. Therefore, the
4 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000

method is applied iteratively, allowing the removal of intersection points among the different routes. To
solve this problem it was developed the following algorithm.

3.1. Algorithm for evacuation route planning

The algorithm proposed is a heuristic algorithm. The basic idea involved is to use interactively an
algorithm for finding the two independent paths in the network from the disaster area (origin) to each
shelter (destination), with the smallest travel time, associating to each path an index C/T (capacity/time)
that will facilitate to evaluate each route comparatively with another. This method will be applied for each
destination point.
For each destination points, the routes found at each iteration of the k-independent routes algorithm
can present coincident nodes, which are removed from the network to verify the existence of other non-
coincident routes in another iteration. The objective is to identify at least two routes which can drain away
the traffic flow from the origin (disaster area) to the destination node (shelter) in the smallest time. In
each iteration, arcs (links), considered “bottlenecks” (arcs of minimum capacity in the path), or coincident
nodes can be removed. The exclusion of bottleneck arcs is made to obtain an improvement in their
capacity and the exclusion of the coincident nodes is to find paths not coincident with the resultant paths
at other interactions. Through several iterations, different routes are found with their respective C/T
indexes. At the end of the process, the best set of two independent routes is chosen.
The k-independent route algorithm used is an existing algorithm proposed by Campos et al. (1999). It
comprehends an iterative process excluding arcs or nodes of the network at each iteration. The nodes and
arcs are removed from two different represented networks, obtained from the original network. In a
represented network, called RG, the bottleneck arcs are removed and, in another, called RN, the
coincident nodes. For finding the two shortest paths in a network, an existent algorithm is used. There are
some algorithms (Shier,1976, 1979; Yen 1971) in the literature that present similar structure. The
peculiarities of the studied problem indicate the use of those algorithms whose results do not present path
with “loops” (paths passing more then once at the same node).
After defining the set of two independent routes, with the best C/T index, from the disaster area to each
destination node (shelter), intersection points that eventually exist among these different sets of route
must be identified. The method is then applied iteratively, allowing the removal of these intersection
points among the different routes. Routes to different destinations points (shelters) cannot have
intersection points either, so their use minimizes the problem of accidents and permits a continuous traffic
flow. However, in cases which a set of routes without intersection points is unfeasible, the method
provides previous knowledge of this intersection point, indicating the need for interventions at this point
in order to avoid potential accidents caused by conflicting movements.

3.2. Structure of the Algorithm

Considering a network R with an origin node S and n-destination nodes Dn, the algorithm has the
following steps:

 Step 1: Identify the area vulnerable to disasters to which the evacuation plan will be applied
 Step 2: Identify the source node S on the network R
The determination of the source node S is made by the use of the center of gravity method in the
potential disaster area.
 Step 3: Define the number of shelters (n) and identify the destination nodes Dn on the network R
 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000 5

 Step 4: Run n times the algorithm for finding k-independent routes from the origin node S to each
destination node Dn, considering k=2
○ Step 4.1: Determine the k-shortest paths
The determination of the k-shortest path between origin O and destination Dn in the network is
made by using an existent algorithm. For each one of the found paths, the total time is obtained:
Tp =  t (a ) p
a p
for p = 1, 2,... ,k (2)

where t(a)p is the travel time in the arcs that belongs to the path p.
If this is the first iteration, go to the step 4.2; otherwise, two situations can happen:
(1) the paths between the origin node S and the destination node Dn does not exist; in this case, if
in the previous iteration arcs were removed (step 4.3), go to step 4.5; if in the previous iteration,
nodes were removed (step 4.5), go to step 4.3. If for two consecutive iterations, there are not
possibilities to find paths, then go to step 4.6.
(2) there are paths between the origin S and the destination node Dn, go to step 4.2.
○ Step 4.2: Identify the bottleneck arcs and define the indexes Cp/Tp
It is verified the maximum flow that can be allocated in each found path, defined by the arc with
the smaller capacity (bottleneck arc of the path, (gp))in the path .
It is defined the capacity Cp = min C(i,j)p and Ip = Cp/Tp (capacity/ travel time index) for each
path p, p=1,.. ,k.
Each passage in step 4.1 defines a iteration that is concluded in the step 4.2. The found paths and
their respective indexes are stored in a list of rk paths, where r represents the number of iterations. In
the first iteration (r=1), the original represented network is used. If the next iteration is an even
iteration, go to step 4.3 for removing bottleneck arcs; if it is an odd iteration, go to step 4.4 for
identification of the coincident nodes.
○ Step 4.3: Modify the network removing arcs
Remove the bottleneck arcs identified in the two previous iterations (or one, if is the second
iteration), and return to step 4.1. This modification is made in the net RG for a reappraisal of the paths
regarding their capacity and the respective indexes Ip.
Be r the current iteration. Then, RGr-2 is the network used in this step in the iteration r-2, from
which the bottleneck arcs were removed. We have:
RGr = RGr-2 - {gr-1p, gr-2p} p = 1,2 ..k (3)
where g p and gr-2p are, respectively, the set of the bottleneck arch of the two previous iterations.
r-1

With the net RGr, go back to step 4.1.

○ Step 4.4: Identify coincident nodes
In this step, it is verified if there are coincident nodes in the paths found until now. If they exist,
verify those that appear with the largest frequency. Be L the number of paths found until the present
iteration, and xl a node of a path l (l =1,2, ..L). It is defined | x l | as the number of paths where the node
xl appears. If | xl |  (L-(k-1), independent k-routes can be found between result paths. In this case, go
to step 4.6; otherwise go to step 4.5.
o Step 4.5: Modify the network removing nodes
Remove coincident nodes from the RN-network, or from the initial network R, if this is the first
passage for this step, and return to step 4.1.
The nodes to be removed are those identified in the previous step. Be XL the group of them and
RNr-2 the net modified by removing the nodes in the previous passage by this step, then make:
RNr= RNr-2 – {XL} (4)
and go to step 4.1.
6 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000

○ Step 4.6: Identification and choice of k-independent route sets

In this step, starting from a group L of paths found in r iterations, it is verified the existence of
independent paths. If they exist, they are put in sets of independent k-paths. Two situations can
happen:
1) there are sets of k paths, in this case the sum of their respective indexes (Cp/Tp) is calculated,
being chosen the set with the largest sum. To decide if it is the best solution, the following
analyses can be made: (i) if there is no more possibility to find paths in both networks (RG and
RN), then the existent solution is the best; (ii) if there is the possibility to find paths, the iterations
are continued and we return to step 4.5. The analysis to identify the possibility to find paths is
based on the concept of potential potent of a node, which is determined by first finding the total
additional flow that can go into each node and the total additional flow that can possibly flow out
of each node and then taking the smaller of the two as the smaller of the two as the potential
potent of the node. This criterion is explained in detail by Campos (1997).
1) there are no sets of k paths,: In this case, there are two possibilities: (i) if the possibility of
finding paths does not exist and a set of paths with less of k elements exist, the found solution is
that which gives the largest sum of indexes among the found sets; otherwise there is no solution;
(ii) if it still exists possibility to find paths, go to step 4.5.
End of the algorithm.
 Step 5: Identify coincident nodes in the n sets of two routes from origin S to destination nodes Dn
If coincident nodes are not found, this is the best possible solution, otherwise go to step 6.
 Step 6: Modify the network removing nodes
Remove the coincident node with the highest rank. If there are two or more coincident nodes with the
same rank, remove the node with the lowest capacity and go to step 4. If the solution is unfeasible, that is,
independent routes were not identified without the coincident node, keep this node and repeat step 6 for
the following coincident node. If all coincident nodes were already analyzed, go to step 7.
 Step 7: Identification and choice of the evacuation routes
Keep the last set of routes as the best evacuation routes. Check for existing coincident nodes. If there
are coincident nodes, plan for its treatment.
End of the algorithm.
Figure 1 presents the steps for the proposed algorithm.

4. Evacuation Planning Exercise

For a better understanding of the developed method and to show its viability, in this section, it is
presented an application of the method as an evacuation planning exercise. In this exercise, a potential
disaster area is defined, as well as potential shelters to which the affected population would be evacuated.
Then the transport network, its capacity and travel time are defined. Figure 2 presents the transport
network used in an Evacuation Planning Exercise in order to enable the evaluation of the model proposed
in section 3, which was computationally implemented in C++. A network composed by 13 nodes and 21
non-oriented arcs and its characteristics (road capacity, flow and travel time) are shown in Figure 2. The
flow of vehicles to be evacuated has origin on node 6 (disaster area), the central point of the potential
disaster area that should be evacuated immediately in the case of the effective event of the disaster. The
destination nodes are nodes 1 and 10, predefined as potential shelters. Among the different possibilities of
evacuation routes in the proposed scenario, the method was able to identify two independent routes from
each shelter (nodes 1 and 10) to the disaster area, considering travel time and capacity of the
transportation network as parameters for analysis. The application of the method is described according to
the steps of the model, presented in Figure 1.
 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000 7

Step 1
Identify areas vulnerable
to disaster

Step 2
Identify source node (centroid of the
disaster area) on the network

Step 3
Define the number of shelters and
identify the destination nodes on the
network
Step 4
Run n times the algorithm for finding k-
independent routes from source S to
destination nodes Dn (k=2)

Step 5
Identify coincident nodes in the sets of
routes from origin S to destination nodes Dn

No Step 7
Coincident
Identification and choice of
nodes?
the evacuation routes
Yes

Step 6 Yes
Modify the network removing
coincident nodes

Fig. 1. Steps for evacuation route planning

10, 1500 9, 2000

1 3 4

4, 1200 5, 1600 6, 1500

6, 1880
2 7
5, 1250
8, 1000
5, 1200
10, 1000 8
6
6, 1580 5, 1500
7, 1670
5 2, 950
9
9, 1050
9, 1450
12, 950

10 7, 1370 11

11, 1870
9, 1380
11, 1250
12 13
Fig. 2. Evacuation Planning Exercise
8 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000

 Steps 1, 2 and 3: Identify the area vulnerable to disasters to which the evacuation plan will be
applied, the source node S and the destination nodes Dn on the network R
Figure 1 shows the road network used for evacuation from the potential disaster area. The
determination of the source node was made by the use of the center of gravity method in the potential
disaster area. Two potential shelters were predefined. The location of the source node (node 6) as well
as of the destination nodes (nodes 1 and 10) were identified in the network, as shown in Figure 1.
 Step 4: Run the algorithm for finding k-independent routes from the origin node S to each
destination node Dn, considering k=2
The k-independent route algorithm (section 3) was used in order to identify 2 independent routes from
node 6 to node 1 and from node 6 to node 10.
○ Finding 2-independent routes from node 6 to node 1
– Step 4.1: Determine the 2-shortest paths
Table 1. The 2-shortest paths in R0
Path order Path nodes Total time

1 6-9-8-2-1 20
2 6-7-8-2-1 22
– Step 4.2: Identify the bottleneck arcs and define the indexes Cp/Tp
The capacity of path 1 is 1200 veic/h, given by the bottleneck arc (2,1), and the capacity of path
2 is 1000 veich/h, given by the bottleneck arc (7,8). Their indexes are, respectively, 60.0 and
45.45. At this first iteration, go to step 4.3 for removing bottleneck arcs.
– Step 4.3: Modify the network removing arcs
Remove from the original network the bottleneck arcs identified in the prvious step and define:
RG1= R0 - {(2,1); (7,8)}
With the net RG1, go back to step 4.1.
– Step 4.1: Determine the 2-shortest paths
Table 2. The 2-shortest paths in RN1
Path order Path nodes Total time

3 6-4-3-1 25
4 6-7-4-3-1 30
– Step 4.2: Identify the bottleneck arcs and define the indexes Cp/Tp
The capacity of path 3 is 1500 veic/h, given by the bottleneck arc (1,3), and the capacity of path
4 is 1200 veich/h, given by the bottleneck arc (6,7). Their indexes are, respectively, 60.0 and
40.0. As in the previous iteration bottleneck arcs were removed, go to step 4.4.
– Step 4.4: Identify coincident nodes
Examining all four paths, we can identify that nodes 2,3,4,7 and 8 appears at least in two
different paths. Each one has | xl | =2 and L=4 (4 paths found until this step), so | xl |  (L-(k-1)
and two independent routes can be found between result paths. In this case, go to step 4.6.
– Step 4.6: Identification and choice of k-independent route sets
 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000 9

In this step, the existence of 3 sets of 2-independent paths is verified.

 I p  105 , 45
K1={2,3}  p  2,3

K2={1,4}   I p  100
p 1, 4

K3={1,3}  p 1,3 I p  120
As the sum of the indexes for set K3 has the highest value, this set is chosen.
Table 3 shows the set of the two-independent routes found.

Table 3. The 2-shortest paths identified from node 6 to node 1

Path nodes Total time Capacity Cp/Tp

6-7-8-2-1 22 1000 45.4

6-4-3-1 24 1500 62.5
The inner potential of the network equals to greatest potential potent of the nodes of the network.
Therefore, the inner potential for the analyzed network (PMI(r)) equals 1500 veic/h. Comparing this
value to the total additional flow that can go into the destination node 1 (PI(T) =1200) and to the total
additional flow that can possibly flow out of the source node 6 (PE(S)=1200), we can verify that the
value of PMI(r) is greater than the minimum of {PE(S), PI (T)}. Therefore, the maximum potential of
the network (PM(r)) equals to 1200 veic/h. Thus, the limit ratio C/T, that is, the ratio of PM(r) and the
highest total travel time among the paths from the last iteration (in this case, T=30 for path 4), equals
to 40, a value lower than the ratios of the independent paths found so far. Therefore, there is no
possibility of finding better routes, so the current solution is optimal.

○ Finding 2-independent routes from node 6 to node 10

The same procedure adopted to find the 2-independent routes from node 6 to node 1 is used in this
case. The final results are shown on Table 4.
Table 4. The 2-shortest paths identified from node 6 to node 10
Path nodes Total time Capacity Cp/Tp

6-9-5-10 21 1450 69.04

6-11-13-12-10 34 1100 32.3

 Step 5: Identify coincident nodes in the 2 sets of two routes from node 6 to nodes 1 and 10
Node 9 is coincident to path 6-9-8-2-1 and path 6-9-5-10, so go to step 6.
 Step 6: Modify the network removing nodes
Node 9 was removed and step 4 was repeated.
10 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000

 Step 4: Run the algorithm for finding k-independent routes from the origin node S to each
destination node Dn, considering k=2

○ Step 4.1: Finding 2-independent routes from node 6 to node 1

The same procedure adopted to find the 2-independent routes from node 6 to node 1, before removing
node 9, is repeated in this case. The final results are shown on Table 5.

Table 5. The 2-shortest paths identified from node 6 to node 1, after removing node 9
Path nodes Total time Capacity Cp/Tp

6-9-8-2-1 20 1200 60
6-4-3-1 24 1500 62.5
○ Step 4.2: Finding 2-independent routes from node 6 to node 10
The same procedure adopted to find the 2-independent routes from node 6 to node 1, before removing
node 9, is repeated in this case. The final results are shown on Table 6.

Table 6. The 2-shortest paths identified from node 6 to node 10, after removing node 9
Path nodes Total time Capacity Cp/Tp

6-9-5-10 21 1450 69.04

6-11-13-12-10 34 1100 32.3

 Step 5: Identify coincident nodes in the 2 sets of two routes from node 6 to nodes 1 and 10
No coincident nodes were found, so this is the best possible solution (Table 6).

5. Conclusion

Evacuation planning is an important part of Emergency Plans, mainly in regions vulnerable to

disasters. The importance of evacuation planning should be emphasized due to its main role in emergency
relief operations. Such plans can assign evacuees to fixed routes or destinations even prior to disasters,
which can help to manage emergencies more effectively. In this context, this paper presents a method for
defining two independent paths from the disaster area (origin) to each destination points (shelters) for
vehicle flow allocation in evacuation planning, considering both travel time and capacity of the
transportation network as parameters for analysis. The routes have no intersection points. However, for
cases in which a set of routes without intersection points is unfeasible, the method provides previous
knowledge of this intersection point, indicating the need for interventions to avoid accidents caused by
conflicting movements. The method identifies two independent paths from the origin to each destination
point, so traffic flow can be allocated in a way that will increase the speed of the evacuation and also
because there will be an alternative route in cases of infrastructure failure or road blocks.
The proposed method was tested in other network cases and its efficacy and usefulness was verified.
The algorithm was computationally implemented in C++ for these tests, even though the program still
needs improvement. Moreover, the network should effectively represent the potential disaster area,
containing the main roads of the area (those that present the best physical conditions for traffic), beyond
the definition of the source (origin) and destination nodes (shelters). The origin of the flow can be only
one and, if there is more than one place (node), it is necessary to create a fictitious origin that will be
linked to the several sources. The proposed method allows a previous analysis of several scenarios,
including: total flow of vehicles in the evacuation, alteration in the orientation of the roads and the
location of the shelter places for the population. It can also be associated to a simulation process.
 Author name / Procedia - Social and Behavioral Sciences 00 (2011) 000–000 11

References

Alexander, D. Principles of Emergency Planning and Management. Oxford University Press, 2002.

Apte, A. (2009.) Humanitarian Logistics: a new field of research and action. Foundations and Trends in
Technology, Information and Operations Management, 3 (1), 1-100.

Barbarosoglu, G. & Arda, Y. (2004). A two-stage stochastic programming framework for transportation
planning in disaster response. Journal of the Operational Research Society, 55, 43-53.

BoaVentura Netto, P. (1996). Grafos: theory, models, algorithms. São Paulo: Publishing Edgar Blucher.

Bollobás, B. (1978). Extremal Graph Theory. New York: Academic Press.

Campos, V.B.G. (1997). Method for flow allocation in transport planning in emergency situations:
definition of disjoint routes, Doctoral Thesis, Eng. of Produção COPPE/UFRJ.

Campos, V.B.G, SILVA, P.; NETTO, P. (1999). Evacuation transportation planning: a method of identify
optimal independent routes. Proceedings of the V International Conference on Urban Transport and the
Environment for the 21st century. Boston: WIT Press.

Chandes, J. & Paché, G. (2009).Investigating humanitarian logistics issues: from operations management
to strategic action. Journal of Manufacturing Technology Management, 21(3), 320-340.

Han, L., Yuan, F., Chin, S. & Hwang, H. Global Optimization of Emergency Evacuation Assignments.
Interfaces, 36(6), 502-513, 2006.

Kovacs, G. & Spens, K. (2007). Humanitarian logistics in disaster relief operations. International Journal
of Physical Distribution & Logistics Management, 37 (2), 99-114.

Kovacs, G. & Spens, K.(2009). Identifying challenges in humanitarian logistics. International Journal of
Physical Distribution & Logistics Management, 39, n. 6, p. 506-528, 2009.

Sapir ,G. (2011). Disasters in numbers 2010. CRED, Catholic University of Louvain, Brussels
(Belgium).

Shier, D. (1976). Interactive methods goes determining the K- shortest paths in network. Networks, 205-
229.

Shier, D. (1979). Algorithms goes finding the k-shortest paths in network. Networks, 195 - 214.

Stepanov, A. & Smith, J. (2009). Multi-objective evacuation routing in transportation networks.

European Journal of Operational Research, 1498, 435-446.

Yen, J. (1971) Finding the K-shortest loopless paths in the network. Management Science, 17 (11).

View publication stats