Applied Soft Computing 27 (2015) 553–566

Contents lists available at ScienceDirect

Applied Soft Computing

journal homepage: www.elsevier.com/locate/asoc

Evolutionary optimization for disaster relief operations: A survey

Yu-Jun Zheng a,∗ , Sheng-Yong Chen a , Hai-Feng Ling b
a
College of Computer Science & Technology, Zhejiang University of Technology, Hangzhou 310023, China
b
College of Field Engineering, PLA University of Science & Technology, Nanjing 210007, China

a r t i c l e i n f o a b s t r a c t

Article history: Effective planning and scheduling of relief operations play a key role in saving lives and reducing dam-
Received 28 December 2013 age in disasters. These emergency operations involve a variety of challenging optimization problems, for
Received in revised form 30 August 2014 which evolutionary computation methods are well suited. In this paper we survey the research advances
Accepted 28 September 2014
in evolutionary algorithms (EAs) applied to disaster relief operations. The operational problems are classi-
Available online 5 October 2014
fied into five typical categories, and representative works on EAs for solving the problems are summarized,
in order to give readers a general overview of the state-of-the-arts and facilitate them to find suitable
Keywords:
methods in practical applications. Several state-of-art methods are compared on a set of real-world emer-
Disaster relief
Emergency operational problems
gency transportation problem instances, and some lessons are drawn from the experimental analysis.
Evolutionary algorithms (EAs) Finally, the strengths, limitations and future directions in the area are discussed.
Optimization © 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-SA
license (http://creativecommons.org/licenses/by-nc-sa/3.0/).

1. Introduction methods are often computationally unaffordable and of too limited

Our world is full of threats from natural and man-made disas-
ters. Particularly, there are an increasing number of heavy disasters
occurred in recent years, such as the 2008 Wenchuan earthquake,
the 2010 Zhouqu mudslides, the 2011 Japan tsunami, and the recent
Haiyan typhoon in Philippines, which have caused serious damage
to lives and property. In a disaster situation, relief operations have
to be planned and implemented effectively in order to be able to
mitigate the damage as much as possible.
From a mathematical viewpoint, disaster relief operations
involve a wide variety of complex optimization problems. The area
of optimization has a vast and fertile overlap between operations
research (OR) and computer science (CS). In general an optimiza-
tion problem can be defined as to find a feasible solution x* ∈ X that
maximizes (minimizes) a objective function f, while X is a finite or
infinite solution space.
Being confronted with an optimization problem, the first con-
cern of computer scientists is whether the problem can be solved
by some efficient algorithms (and in particular in polynomial time).
However, most problems arising in OR are proved to be NP-hard,
i.e., polynomial-time intractable unless P = NP [1]. Moreover, under
emergency conditions the solutions to operational problems need
to be developed within a very limited time. Hence, traditional exact
applicability.
Evolutionary computation is a class of optimization methods
drawing inspiration from natural evolution and adaptation [2]. Due
to their simplicity, sufficient flexibility and general applicability,
evolutionary algorithms (EAs) are well suited to tackle a wide range
of computationally intractable problems. They do not always guar-
antee reaching the exact optimal solution in a single simulation
run, but in most cases they can obtain satisfying solutions within an
acceptable computational time. Particularly, intrinsic probabilistic
behaviors render them capable of effectively handling real world
problems involving nonlinearity, complexity, noisy environment,
imprecision, uncertainty and vagueness [3,4]. Consequently, evo-
lutionary optimization has attracted wide attention and has been
extensively explored in OR in recent decades.
The research on evolutionary optimization methods for disas-
ter relief operations, in comparison with other sub-areas of OR, is
limited. But it has started to evolve a lot in the last years. A variety of
EAs, including genetic algorithm (GA) [5], particle swarm optimiza-
tion (PSO) [6], ant colony optimization (ACO) [7], etc., have found
a growing number of applications in relief operations (as shown in
Fig. 1), and demonstrated their effectiveness on many challenging
emergency problems.
In this paper we present a survey of research works on this topic,
the main purpose of which is threefold:

∗ Corresponding author. Tel.: +86 571 85290085. 1. Provide readers a general overview of the major developments
E-mail address: yujun.zheng@computer.org (Y.-J. Zheng). emerged throughout the years.

http://dx.doi.org/10.1016/j.asoc.2014.09.041
1568-4946/© 2014 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-SA license (http://creativecommons.org/licenses/by-nc-sa/3.0/).
554 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566

target j (1 ≤ i ≤ m, 1 ≤ j ≤ n), such that the total transportation cost is

minimized. It can be modeled as a special case of the linear program
(LP) as follows:


m

n

min f = cij xij

i=1 j=1

m
s.t. xij ≥ bj , j = 1, . . ., n
(1)
i=1
n
xij ≤ ai , i = 1, . . ., m
j=1
xij ≥ 0, i = 1, . . ., m, j = 1, . . ., n

Fig. 1. The numbers of research articles on EAs for disaster relief operations in recent
ten years (the data of 2014 is up to May).
2. Show major strengths and shortcomings of the state-of-the-arts,
and help practitioners to find valuable approaches that can be
referred in the practice of disaster relief.
3. Discuss potential directions for future research, and stimulate
more interest in this cross-disciplinary field.
In the remainder of the paper, we first classify optimization
problems in disaster relief operations into five typical categories
in Section 2, and describe recent advances in EAs for solving the
problems in Section 3. Section 4 presents the experiment of six typ-
ical EAs on a set of real-world emergency transportation problem
instances. Section 5 discusses the strengths, limitations and future
directions in the area, and finally Section 6 concludes.
Although this problem can be solved by specialized methods
that are more efficient than standard linear programs [10], EAs can
provide much better performance on very large problem instances
[11]. Moreover, real-world problems typically have to deal with
heterogenous commodities and transportation modes, involving
additional (and often nonlinear) objectives and constraints [12].
Such problems should be characterized by more complex integer
programming (IP) or mixed integer programming (MIP) models,
which are much better suited for evolutionary optimization than
standard optimization method.
Furthermore, transportation in emergency logistics [13] is often
subject to uncertainty and randomization, has very limited time
constraints, focuses more concern on timeliness and/or risk rather
than transportation cost. Taking these extra properties into consid-
eration, much more efforts should be put into the design of efficient
algorithms.

2. Problem classification
2.2. Location problems
Disaster relief operations involve activities including estab-
lishing emergency facilities, searching and rescuing survivors,
The problems of locating emergency facilities, such as fire sta-
providing health and medical assistance, distributing relief sup-
tions, medical services, shelters, etc., based on the topography of
plies, transferring injuries, scheduling rescue forces, etc., and the
potential facilities, can be divided into two sub-classes: continu-
coordination of these activities across organizations [8,9]. Thus they
ous facility location problems, where facilities are allowed to be
are associated with a variety of operational problems that often lie
sited on every point in a planning area, and discrete facility loca-
outside the scope of the conventional optimization methods. In this
tion problems, where there is a discrete set of candidate locations.
paper, we classify the problem into the following five categories:
Generally speaking, the first class often uses formulation of LP and
nonlinear programming (NLP) problems, while the second class is
• General transportation planning problems, which are to make up
often modeled by classical covering problems and their variants,
programs for delivering relief supplies from distribution centers such as set covering, vertex/egde covering, covering location, and
(sources) to demand points (targets). But we do not place detailed partial covering [14,15].
path planning and vehicle routing in this category. The simplest form of the facility location problem, which con-
• Facility location problems, which are to arrange emergency facil-
siders providing a single facility to cover a maximum number of
ities on appropriate locations to serve the demand points. demand points, has efficient exact solution methods [16,17]. How-
• Routing problems, which include planning routes for vehicles,
ever, involving multiple facilities leads to NP-hard problems, such
rescuers, and evacuees. as the p-median problem (2) and the maximum covering problem
• Roadway repair problems, which are to repair damaged roadways
(3):
and rehabilitate the lifelines to demand areas.
• Integrated problems, which need to solve a set of above individual

problems under one or more common objectives. min f = cij yij

2.1. General transportation planning problems  i∈D j∈F

s.t. xj = p
As a classic problem in OR, the basic transportation problem 
j∈F
(2)
considers delivering a homogeneous commodity from a set of m yij = 1, ∀i ∈ D
sources to a set of n targets. Suppose the supply of source i is ai , j∈F
the demand of target j is bj , the cost for transporting one unit of xj ∈ {0, 1}, ∀j ∈ F
commodity from source i to target j is cij , then the problem is to
determine the commodity amount xij from each source i to each yij ∈ {0, 1}, ∀i ∈ D, ∀j ∈ F
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566

max f = wi zi

i∈D
s.t. zi − xj ≤ 0, ∀i ∈ D
(3)
j∈F
xj ∈ {0, 1}, ∀j ∈ F
zi ∈ {0, 1}, ∀i ∈ D
where F is the set of candidate facility locations, D is the set of
demand points, and cij is the cost of assigning demand point i to
facility j.
Moreover, under disaster environment, facility location often
needs to deal with limited facility resources, tremendous demands,
and high pressures, the combination of which can lead to uncertain
and insufficient services for demand points. Thus the formulation
of real-world emergency facility location problems, especially in
large-scale operations, should take into consideration redundant
facilities, service priorities, as well as location dynamics [15], to
ensure an acceptable demand coverage.
2.3. Routing problems
We also divided this category into two classes: long-distance
(vehicle) routing, and short-distance routing for rescuers and evac-
uees. Both classes are typically defined on a graph G = (V, E), where
V is the vertex set and E is the edge set, and the problems are to
search for one or a set of routes (paths in the graph) that opti-
mize one or more objective functions whilst satisfying all given
constraints.
Vehicle routing problems (VRP) are a class of fundamental NP-
hard problems which have been widely studied in the literature
[18]. VRP in emergency logistics are characterized as scheduling
limited vehicles to accomplish delivery tasks in rigidly limited
times and under harsh environment conditions. Due to the large
number and various types of supplies to be delivered, emergency
VRP typically involves multiple depot, trips, and heterogeneous
vehicles, and thus much more difficult than VRP and their variants
in common supply chains.
Short-distance routing mainly concerns with planning effective
evacuation or search-and-rescue paths, which are vital for saving
lives and reduce losses in sudden events. In contrast to ordinary
path planning problems which are to find a shortest or k-shortest
paths in a weighted graph, emergency evacuation/rescue path
planning should take more complex properties, such as obstacle
avoidance, path following, and congestion control, into consider-
ation. And in most cases, response speed is more important than
the optimality of the plans.
2.4. Roadway repair problems
Emergency roadway repair is to efficiently schedule limited
repair resources to rehabilitate the roadway network, sufficient
to carry out other relief operations. Ordinary shop job scheduling
models and solution methods [19] generally do not apply, because
they fail to capture time dependencies among repair operations,
e.g., many damaged points are unreachable unless some other dam-
aged points have been repaired. By incorporating time variables,
Yan and Shih [20] developed a time-space network that models
the roadway repair problem as an integer network flow problem
with side constraints. However, a repair schedule with the mini-
mum length of time is not always the most efficient in supporting
subsequent relief operations [21], and more comprehensive crite-
ria are required for properly evaluating solutions to the complex
problem.
555
3. Evolutionary optimization methods for relief
operational problems
As mentioned, due to the rising complexity and stringent
response time requirements of today’s relief operational problems,
conventional exact solution methods are generally not applicable.
In comparison, EAs have advantages in high efficiency in exploring
extra-large solution spaces, and inherent ability to handle uncer-
tainty and randomness, robustness and insensitivity to noises, and
scalability to high dimensions, and thus are more much suitable for
solving these emergency problems.
In particular, emergency relief operational problems often
involve multiple objectives, such as minimum operational time,
maximum rescue effectiveness, minimum operational risk, etc.,
which can conflict with one another. For such multiobjective
optimization (MOO) problems, it is preferable to simultaneously
optimize all the objectives by evolving a set of solutions to the so-
called Pareto front to provide a comprehensive decision support
[22,23]. Multiobjective EAs (MOEAs) are favorable for this purpose,
because they concurrently evolve a population of possible solutions
and thus enable to find several members of the Pareto optimal set
in a single run of the algorithm, instead of having to perform a series
of separate runs as in the case of traditional optimization methods
[24].
In this section, we describe previous studies in the literature
on evolutionary optimization methods for emergency relief opera-
tional problems. Particularly, on each method, we identify whether
it has the following characteristics:
• Unc/Rand – The method deals with problems with uncertainty
and randomness.
• MOO – The method deals with MOO problems.
• PBO – The method employs a Pareto-based optimization proce-
dure for solving MOO problems, instead of transforming multiple
objectives into a single objective.
• M-Pop – The method uses multiple populations to cooperatively
solve the problems, which extends the basic single-population
based EAs to co-evolutionary methods [25].
Also, for the convenience of readers, Appendix (Table 9) clas-
sifies the methods according to the five categories described in
Section 2.
3.1. Genetic algorithms
GAs are the most well-known members of EAs which simulate
the Darwinian principle of natural selection and the survival of the
fittest in optimization. Typically, a GA works with a fixed-size pop-
ulation, uses genetic operations including crossover and mutation
to evolve the solutions, and selects the most appropriate offspring
to pass on to next generations. Table 1 summarizes the character-
istics of GA-based approaches for emergency operational problems
in the literature.
3.1.1. GAs for general transportation planning
An early work of Tuson et al. [26] developed a prototype system
of emergency resource distribution planning on a 386-class PC. The
basic problem of the system considers that there are n sites, each of
which has a minimum demands of m resources, and shortages and
surpluses of each source occur at different sites. The aim is to max-
imize the number of resource target mets whilst minimizing the
number of shipments. The system employs several heuristic meth-
ods, mainly based on GA and its variants. A solution is encoded as
the order of sites by which their shortages are to be tentatively sup-
plied, and the crossover operator used is ‘modified PMX’ [53] which
has been found to give good results on many permutation-encoded
556 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566

Table 1 represented with genes composed of delivery points and products.

The characteristics of GA-based approaches.
The algorithm employs the set enumeration heuristic to initialize
Unc/Rand MOO PBO M-Pop a population of solutions, and uses three fast improvement proce-
√ dures to evolve the solution, replacing some worst solutions with
Tuson et al. [26]
√
Feng and Wen [27] new ones at each generation for diversification. Computational
Na and Zhi [28]
√
experiments show that the method can produce near optimal solu-
Nolz et al. [29] tions in relatively short computation times and is fast enough to be
√ √
Yang et al. [30]
used interactively in disaster response.
Berkoune et al. [31]
√
Wang and Zhang [32]
Kongsomsaksakul et al. [33]
√ 3.1.2. GAs for facility location
Jia et al. [34]
Han and Zhang [35] For optimizing the location of catastrophic rescue centers, Wang
Chai et al. [36]
√ √
and Zhang [32] presented a model taking account of the probabil-
Tzeng and Chen [37] ity of catastrophic occurrence, catastrophic diffusion function and
√ √
Yang et al. [38]
rescue function, and developed a GA using binary encoding and
Ma et al. [39]
Lin et al. [40] embedding a greedy heuristic for fitness calculation. Experiments
√
Zhang et al. [41] on examples arising practice show that the algorithm provides sat-
√
Hamedi et al. [42]
√ √ isfactory results.
Chung et al. [43] Kongsomsaksakul et al. [33] studied the optimal shelter loca-
Liu et al. [44]
√ tions for flood evacuation planning. They formulated the problem
Song et al. [45]
√ √ using a bi-level programming model, with the upper level for loca-
Saadatseresht et al. [46]
√
Ma et al. [47] tion and the lower level for traveler behavior. They developed a
Sato and Ichii [48] GA that uses genetic operations to evolve a population of complete
√ √
Chen and Tzeng [49]
√ solutions, and employs a double-stage algorithm to evaluate the
Zhang and Lu [50]
Chang et al. [51] fitness value of each chromosome.
√ √ √ √
Zheng and Ling [52] Considering the facility locations of medical supplies, Jia
et al. [34] proposed a maximal covering problem model, where
the demand uncertainty and medical supply insufficiency are
addressed by providing each demand point with services from a
problems. But the experiments indicated that the GAs perform less
multiple quantity of facilities that are located at different quality
well than the stochastic hill-climbers with the iteration.
levels (distances). Three heuristics including a GA are developed
Feng and Wen [27] used bi-level programming to construct
for the location problem, and the experiments show that the GA is
three models for traffic control in disaster raided area. The basic
appropriate for small problems but produce less-quality solutions
model is to maximize traffic volume, the second model takes both
on large problems (mainly because of premature convergence).
the needs of rescue and non-rescue vehicles into consideration,
Han and Zhang [35] studied an extended emergency facility
and the last model further integrates the combined distribu-
location problem which is modeled as an IP problem, and proposed
tion/assignment model to meet the lower level objectives. A GA
a GA where each chromosome is constituted by the serial num-
with redefined crossover and mutation operators is applied to seek
ber of each target’s emergency facility, and population diversity
the approximate optimal solution to the problem.
is maintained by using a changeable mutation probability. Experi-
Na and Zhi [28] constructed an emergency relief goods distribu-
ments show that the GA solve the problem much more efficiently
tion model which considers multiple transportation modes. Given
than other simple heuristics.
different weights to the different modes, the objective of the model
Chai et al. [36] studied an highway traffic emergency facil-
is to minimize the total weighted transport time. They adopted a GA
ity location problem, and proposed a heuristic GA which uses an
to solve the problem, where natural number coding, tournament
n-dimensional 0–1 integer vector to present chromosomes, and
selection and single-point crossover are used, and a penalty func-
employs a heuristic operator for improving unreliable chromo-
tion is included in the fitness of individuals for handling problem
somes by (1) remedying the chromosomes that do not cover all
constraints.
demands and (2) eliminating the redundancy of chromosomes that
Nolz et al. [29] modeled a disaster relief operation planning
cover the location set with redundancy. The method has been
problem with two objectives including the maximal covering loca-
successfully applied to problems from the highway network of
tion and minimum tour length, and they developed a hybrid
Nanjing, China.
method based on NSGA-II [54] with variable neighborhood search
In order to determine the optimal number and locations of fire
and path relinking. Test on real world data from Ecuador demon-
stations at an international airport, Tzeng and Chen [37] proposed
strated the effectiveness of the proposed approach.
a fuzzy multiobjective model, the objectives of which include min-
Yang et al. [30] investigated a railway freight transportation
imizing the total setup cost and minimizing the longest distance
planning problem under the mixed uncertain environment of fuzzi-
from the fire stations to any incident point. The objectives are con-
ness and randomness. Based on the chance measure and critical
verted to a single unified ‘min-max’ goal, and a GA is employed for
values of random fuzzy variables [55], they constructed three
effectively solve the problem. Yang et al. [38] extended the model
chance-constrained programming models of the problem, and
such that different risk categories and obstacles within a given
proposed a hybrid algorithm including potential path searching,
region are considered in the objectives and constraints, and adapted
simulation, and GA, for solving the models to get an optimal solu-
the GA by embedding the constraints into the fitness function.
tion with the least total relevant cost. Numerical examples showed
that the algorithm is steady and robust for not very large scale
problems. 3.1.3. GAs for routing
Berkoune et al. [31] formulated a practical emergency trans- For path optimization of single vehicle in emergency supplies
portation problem with not only multi-source and multi-target, transportation, Ma et al. [39] developed a hierarchical GA method:
but also multiple product types and vehicle types. To solve the Local optimization is achieved by Bottom GA in subnetwork, global
complicated problem, they developed a GA where an individual is optimization is achieved by top GA in the whole network, and the
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
contradiction of the two levels is solved by maintaining a balance
between GA random search and narrow search.
Considering a much more complex model with multi-items,
multi-vehicles, multi-periods, soft time windows, and a split deliv-
ery strategy, Lin et al. [40] formulated a multi-objective integer
programming model and developed two heuristic approaches: The
first is based on GA, while the second approach further decomposes
the original problem. Computational study shows that the GA can
provide high-quality solutions under complex conditions.
Zhang et al. [41] presented a multi-objective split delivery VRP
(SDVRP) in multi-period emergency logistics, the aim of which is to
minimize the unmet demand, total delivery time, and unbalanced
supply among demanders. An aggregated objective is defines to
compromise the three objectives, and a modified GA is proposed
for solve the SDVR. In addition, a clustering method is embedded
for removing infeasible or bad chromosomes so as to promote the
population quality.
Paying attention to reliable route choice of humanitarian
response planning for disaster response, Hamedi et al. [42] for-
mulated the routing and scheduling of supply transportation as a
linear integer programming (LIP) model, and proposed a GA-based
heuristic method for optimizing the source-target pair ordering
and ranking, where a time-dependant shortest path approach is
embedded for chromosome evaluation. The model and heuristic
are flexible in accommodating multiple commodity and regulation
parameters, i.e., each truck can have its own risk time window or
parameter sets.
Chung et al. [43] presented a multiobjective dynamic length GA
to search vehicle routes to deliver the supplies. In the algorithm, a
chromosome be divided into several sections so that the number
of vehicles can be dynamically adjusted, and a domination-based
ranking is used for schedule evaluation and tournament selection.
Liu et al. [44] used a bi-level optimization model for optimal
evacuation planning: the high level tries to maximize the through-
put in the target evacuation clearance time, whereas the low level
is to minimize total clearance time. They adopted a GA to solve
the problem, and the result demonstrates better evacuation time
in comparison to standard design plan.
Song et al. [45] studied the transit bus operations for emergency
evacuation, which is formulated as a location-routing problem with
uncertain demands. They designed a new multi-graph street net-
work considering prohibited turns and intersection delays, and
proposed a hybrid intelligent algorithm including GA, artificial
neural network (ANN) and hill climbing to solve the proposed
stochastic programming model. The algorithm employs partially
matched crossover, route mutation and depot mutation operations.
The application of the method in transit evacuation planning is
illustrated by a practical example of partial Gulfport, USA.
Saadatseresht et al. [46] proposed an MOO model for evacua-
tion planning in three steps, namely safe area designation, optimum
path finding, and optimal safe area selection. They investigated dif-
ferent algorithms for solving the spatial MOO problem, and found
that NSGA-II can obtain the most proper patterns.
To help the injured reach the destination safely and quickly
in a sudden event, Ma et al. [47] receptively presented a
chance-constrained programming model and a chance-dependent
constraint programming model for emergency evacuation rout-
ings under fuzzy environment, and proposed a hybrid intelligence
algorithm to solve the models. The algorithm represents solutions
by produces chromosomes based on priority-based coding, uses
genetic operations to evolve the chromosomes, and integrating
simulation technique to evaluate the chromosomes.
3.1.4. GAs for roadway repair problems
Sato and Ichii [48] conducted a study on the optimization
of restoration process of damaged lifelines by earthquakes, and
557
Table 2
The characteristics of ES/EP-based approaches.
Unc/Rand MOO PBO M-Pop
√
Xie and Zhang [59]
Barrachina et al. [60]
√
Wang and Jin [61]
developed a simple GA and a hybrid GA to decide the priority
of component to restore. Numerical examples show that the lat-
ter performs better than the former. They also applied a single
populated GA to distribute restoration teams at damaged sites for
optimizing the restoration process of lifeline networks.
Chen and Tzeng [49] established a fuzzy multi-objective plan-
ning model for post-quake road-network reconstruction. The
model is formulated as a bi-level network design problem: The
upper level is for work scheduling of many worktroops, and the
lower level is for asymmetric traffic assignment. They converted
the problem to maximize the satisfying degree (or achievement
level) among constraints of conflicting objectives, and developed
a modified GA to obtain a heuristic solution that approximately
optimizes the objectives on the two levels.
Zhang and Lu [50] also proposed a multi-objective model of
emergency roadway repair, the aim of which is to minimize the
length of time as well as the risk of repair. They developed a GA
which shows satisfying performance on the test problems. Nev-
ertheless, how to appropriately and quantitatively assess the risk
remains a difficult challenge.
3.1.5. GAs for integrated problems
Chang et al. [51] sought to generate transportation plan and
vehicle routing schedules, in order to simultaneously minimize
unsatisfied demand, total transportation time and total transporta-
tion cost. They proposed a multiobjective GA that initializes a
population of solutions matching the relief resources between
nearby sources and targets greedily, performs crossover operations
based on partial map reverse exchanging with repair, and ranks the
solutions based on the nondominated sorting of NSGA-II [54]. The
proposed approach was applied to the case of the Chi–Chi earth-
quake in Taiwan to verify its performance.
Zheng and Ling [52] proposed a multiobjective fuzzy opti-
mization problem of emergency transportation planning, which
integrates task and resource allocation at the strategic level, and
delivery scheduling and vehicle routing for individual sources.
Given that many parameters are described as fuzzy variables, the
ˇ-dominance relation and three correlated fuzzy ranking criteria
are employed for evaluating the solutions. They developed a coop-
erative optimization method that divides the integrated problem
into a set of subproblems, uses multiobjective GA to evolve the
sub-populations of solutions to the subproblems concurrently, and
brings together the sub-solutions to construct complete solutions.
Experiments show that the proposed method is robust and scalable,
and outperforms other state-of-the-art MOEAs.
3.2. Evolution strategies and evolutionary programming
Evolution strategies (ES) [56] and evolutionary programming
(EP) [57] are two main variants of GAs. Both ES and EP primarily
use mutation to produce new solutions and use a selection scheme
to test which solutions should survive to the next generation. The
classical EP conducts pairwise comparison over the union of parents
and offspring and the winners enter into the next generation. In
comparison, the state-of-the-art ES is ( + )-ES, which means the
 parents generate  offspring, and the best  solutions form the
next generation [58]. Table 2 summarizes the characteristics of ES
and EP based approaches for emergency problems in the literature.
558 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
Table 3
The characteristics of PSO-based approaches.
Unc/Rand MOO PBO
√ √
Uno et al. [64]
√
Chai et al. [65]
Pang et al. [66]
√
Bozorgi-Amiri et al. [67]
Gan et al. [68]
√ √
Tian et al. [69]
√ mulated nonlinear IP, they proposed a modified PSO algorithm
Wang et al. [70]
√ √
Zheng et al. [71]
Considered a multi-objective decision-making model for min-
imizing both the total losses caused by the fire and the
commencement time of the fire activity, Xie and Zhang [59] pro-
posed a dual-gene variation ES algorithm which takes a ( +  + )
strategy. That is, at each generation,  individuals are selected for
variation from a population of  solutions, and then the best indi-
vidual of the parents carries out the Gauss variation and the worst
individual carries out the Cauchy variation, and finally the best 
individuals among all the parents and children are chosen for the
next generation. The proposed strategy is shown to be better than
other typical strategies and single-gene variations for the emer-
gency problem.
Barrachina et al. [60] considered four different versions of VRP,
with the aim of minimizing the arrival time of the emergency ser-
vices whereas the rest of vehicles are not significantly affected. They
proposed several approaches including a basic ES and a density-
based ES, and tested them on cases with different topologies and
traffic densities. Results show that the density-based ES can achieve
the best performance.
Wang and Jin [61] presented a multi-resource emergency
resources scheduling problem, the aim of which is to minimize
both the total time and the total cost of scheduling. Their solution
method converts the two objectives into a single objective by aggre-
gation, includes penalty to handle constraint violations, and adopts
an improved EP algorithm to optimize the problem. Experiment
shows the proposed method is superior to the basic EP, GA, and
differential evolution (DE) [62] in convergence speed and solution
accuracy.
3.3. Particle swarm optimization
PSO [6] is also a population-based global optimization method
that uses a number of individual solutions, called particles, to
explore a hyper dimensional search space. Each particle has a posi-
tion vector and a velocity vector, which are adjusted at iterations
by learning from a local best found by the particle itself and a cur-
rent global best found by the whole swarm. Empirical studies have
shown that PSO has a much higher efficiency in convergence to
desirable optima than many other EAs on unimodal fitness land-
scapes [63]. Table 3 summarizes the characteristics of PSO-based
approaches in the literature.
Uno et al. [64] proposed a multiobjective programming model of
optimal emergency facility location with two objectives: one is to
minimize the maximal distance from emergency facilities to hos-
pitals, and the other is to maximize frequency of accidents that
emergency facilities can respond quickly. They applied an inter-
active fuzzy satisfying method with PSO for searching an optimal
solution for each of the min–max problems with the corresponding
reference membership values.
Chai et al. [65] established a stochastic programming model
of traffic emergency resources allocation on expressway, with
stochastic variables in constraints and the objective to mini-
mize the overall allocation cost. They developed a discrete PSO
using a presentation of indirect particle position and performing
stochastic simulation to check feasibility after updating particle
positions. The model and the algorithm are used in the case of
M-Pop rescue resource allocation on expressway networks in Nanjing area.
Based on the analysis of practice of emergency management
in China, Pang et al. [66] presented a distribution model of emer-
gency materials in a three-layer relief transportation network, the
aim of which is to minimize the system losses while meeting the
constraints of emergency response time and fairness. For the for-
where different learning objects are used for different dimensions
of the particles.
Bozorgi-Amiri et al. [67] studied a disaster relief logistics
problem where demands, supplies, cost of procurement and trans-
portation are considered as uncertain parameters. Therefore, the
problem is modeled as an uncertain, nonlinear MIP. They used the
robust optimization technique [72] to handle the uncertainty, i.e.,
describing uncertain parameters by the discrete scenarios or a con-
tinuous range. And they designed for the problem an improved
PSO algorithm which combines discrete and continuous compo-
nents in particles. Experimental results show that the proposed
algorithm performs much better than the traditional branch-and-
bound (B&B) algorithm.
Gan et al. [68] studied an emergency VRP with an exponential
utility function designed as an indicator of operational efficiency.
In their PSO algorithm, an n + k − 1 dimensions encoding scheme is
applied to a problem with n disaster areas and k vehicles, and the
objective value is adjusted by penalty function method so that the
constrains are mirrored.
Tian et al. [69] considered a multiobjective emergency trans-
portation model that integrates supplies distribution and vehicle
routing. They designed a PSO algorithm where every particle is
encoded as a discrete-continuous vector, and the discrete vari-
ables and continuous variables are respectively changed using their
own velocity/location update formulas. The fitness function for
evaluating particles is defined by combining multiple objectives.
However, for large-size instances the algorithm is sometimes easy
to be trapped by local optima.
Wang et al. [70] proposed a novel emergency evacuation model
by using linear weight decreasing PSO to simulate individual’s
movement in evacuation. The model has been implemented in an
emergency evacuation simulation system based on GIS, and well-
performed some typical evacuation behaviors such as avoiding
impact, queuing, congesting and herding, in comparison with social
force model and cellular automata model.
Zheng et al. [71] proposed a multiobjective PSO for popula-
tion classification in fire evacuation operations, with the aim to
simultaneously optimizes the precision and recall measures of
the classification rules. The algorithm enables every particle to
potentially learn from a different exemplar at each dimension, and
thus improve the search performance meanwhile maintaining the
swarm diversity. The approach has been successfully applied to a
real-world fire evacuation operation occurred in Zhejiang Province,
China.
3.4. Ant colony optimization
ACO [7] mimic the behavior of real ants living in colonies that
communicate with each other using pheromones in order to find
the shortest path to the food source. In comparison with other
EAs, ACO is specially suitable for path planning and routing prob-
lems. However, it often requires more efforts to extend existing
ACO solvers to path planning in the complex context [73]. Table 4
summarizes the characteristics of ACO-based approaches in the
literature.
Lim et al. [74] considered routing of unmanned reconnaissance
aerial vehicles on sparse graphs that mimic inaccessible zones such
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
Table 4
The characteristics of ACO-based approaches.
Unc/Rand MOO PBO
Lim et al. [74]
√ √
Yuan and Wang [75]
√ √
Yi and Kumar [76]
Yan and Shih [77]
as those hit by forest fires, earthquakes, or other disasters. In addi-
tion to minimize the travel cost, the problem also concerns the
location of low risk paths which can be represented based on a
Voronoi diagram. They developed a hybrid ACO algorithm which
incorporates with local search heuristics for uncovering feasible
routes within tractable times. Empirical results demonstrate the
excellent convergence property and robustness of the algorithm in
uncovering low risk and Hamiltonian visitation pathes.
Yuan and Wang [75] also used ACO to solve two path selection
problems in emergency logistics management. The first is a single-
objective model to minimize total travel time along a path. The
second is a multi-objective model to minimize the total travel time
as well as the path complexity, and it further consider the chaos,
panic and congestions in time of disaster. After converting multi-
ple objectives into a single objective using weighted aggregation,
the proposed ACO algorithm shows good performance on the test
problems.
Considering an emergency logistics planning model which
involves dispatching commodities to distribution centers and evac-
uating the wounded people to medical centers, Yi and Kumar [76]
proposed an improved ACO algorithm that decomposes the prob-
lem into two phases, namely the vehicle route construction and
the multi-commodity dispatch, and then iteratively tackles the
sub-problems. The first phase builds stochastic paths under the
guidance of pheromone trails, and the second phase employs a
network-flow solver for the assignment between different types
of vehicle flows and commodities. As indicated by the experi-
ments, the ACO can achieve quality solution within a minute of
runtime which is acceptable for the planner in real emergency
situation.
Yan and Shih [77] developed a hybrid algorithm based on ant
colony system coupled with a threshold accepting technique, for
solving a time-space network flow problem which is to minimize
the length of time needed for emergency repair of highway network
Yan and Shih [20]. Since a solution to the problem includes sev-
eral routes with a schedule and the arc connections in the roadway
network change with the repair process, generation of feasible solu-
tions (ants) the algorithm is significantly different from the that for
solving TSP. Thus the algorithm sets a pheromone value on each arc
according to a work time limit for every work team, uses the label
correcting algorithm to find the spatial shortest path between every
pair of points, and uses a path-and-time tracing method to compute
the detailed routes and schedules. Test problems using different
roadway network patterns and a case study based on data from the
1999 Chi–Chi earthquake in Taiwan demonstrate the robustness of
the model and the algorithm.
3.5. Biogeography-based optimization
Biogeography-based optimization (BBO) is a relatively new evo-
lutionary algorithm inspired by the science of biogeography [78].
In the emergency response management system proposed by Iyer
and Shanthi [79], BBO was employed to compare and cluster
satellite images, in order to capture accident spots and evalu-
ate their severities. Zheng et al. [80] developed a hybrid BBO for
an emergency railway wagon scheduling problem that consid-
ers multiple source, target, and central stations. The algorithm
559
Table 5
The characteristics of BBO-based approaches.
M-Pop Unc/Rand MOO PBO M-Pop
Iyer and Shanthi [79]
Zheng et al. [80]
√ √
Zheng et al. [81]
Zheng et al. [82]
includes DE mutation to enhance the exploration ability and takes
some problem-specific mechanisms for fine-tuning the search. In
[81] the authors proposed a multiobjective BBO algorithm for an
emergency engineering rescue scheduling problem, which involves
multiple rescue teams and tasks and different and perhaps fuzzy
processing times. The algorithm employs the fast nondominated
sorting method to evaluate the habitat’s suitable index (HSI) of each
solution, and defines new migration and mutation operators for
solution evolution. The method exhibits competitive performance
on the test problems from disaster relief operations occurred in
China in recent years, and has been successfully applied to the 2013
Dixi Earthquake.
Zheng et al. [82] further developed an ecogeography-based
optimization algorithm (EBO), which enhances BBO with a local
topology, and uses the combination of a global migration and a
local migration to perform effective exploration and exploitation
in the solution space. The new algorithm has been applied to an
emergency airlift problem in the 2013 Lushan Earthquake, China.
Table 5 summarizes the characteristics of BBO-based
approaches in the literature.
3.6. Artificial immune system
Artificial immune system (AIS) is a model stimulating the adap-
tive immune system of a living creature to unravel the various
complexities in optimization problems [83]. Hu [84] presented
an immune multi-affinity network model of emergency logistics
network and defined affinity measures to represent complex col-
laborative relationship among emergency logistics components.
Based on the affinity network, the author studied a container multi-
modal transport problem in emergency relief that is to maximize
time efficiency meanwhile minimize the cost, and proposed a solu-
tion approach that combines two objectives into one by using
weight aggregation and goal programming, and then uses tradi-
tional integer linear programming method to find the optimal path
for emergency relief [85].
By extending the AIS algorithm for two-dimensional robot
path planning [86], Liu and Zhang [87] developed an enhanced
AIS algorithm for dual aircrafts path planning for forest fire res-
cue in three dimensional space. The problem considers aircrafts
taking off at different sites and arrival at the destination at the
same time. The algorithm defines an affinity function for eval-
uating the threats to different flight pathes, and employs three
crossover operators for diverse the search. Results show that the
method is capable of planning an efficient flight path in complex
environments.
Zhang et al. [88] proposed a novel bio-inspired algorithm which
is based on the mathematical model of amoeboid organism, to
solve the optimal route selection problem in disaster extension.
The problem considers both the travel time and the path length.
The algorithm uses the kth shortest path program to find out the
1st to kth shortest path, uses the amoeboid algorithm to find the
longest path to construct the dimensionless indexes, and finally
obtains the optimal routes.
Table 6 summarizes the characteristics of AIS-based approaches
in the literature.
560 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
Table 6
The characteristics of AIS-based approaches.
Unc/Rand MOO PBO
√
Hu [85]
Liu and Zhang [87]
√
Zhang et al. [88]
3.7. Other metaheuristic algorithms
Other heuristic search methods including simulated annealing
(SA), tabu search (TS), harmony search (HS), variable neighborhood
search (VNS), etc., are effective particularly in combinatorial opti-
mization problems, and have also been reported in the literature
in for disaster relief. SA [89] is a classical global search heuris-
tic that simulates the physical annealing process in the field of
optimization. Jahangiri et al. [90] presented a problem for opti-
mizing signal timing and increases the outbound capacity of the
network for emergency evacuation, and applied a SA algorithm
for searching the optimal solution to the problem. Ma et al. [91]
designed a genetic-SA algorithm for emergency service facility
location problem. Experiments showed that algorithm can clearly
gets the afflicted points that are covered by the selected facilities
and thus improve searching efficiency.
TS uses an adaptive memory named tabu list to record the vis-
ited solutions and thus improves the local search to avoid local
optima [92,93]. Xie and Turnquist [94] solved a lane-based evacu-
ation network optimization problem that integrates lane reversal
and crossing elimination strategies by using a relaxation and TS
algorithm, and applied the approach to optimize a regional evacu-
ation network for a nuclear power plant. Zheng et al. [95] designed
a multiobjective TS for an emergency equipment maintenance
scheduling problem, which aims to achieve a good balance between
operational capability, reserved maintenance capability, as well as
cost-effectiveness. The approach has demonstrated its effective-
ness in several real-world operations.
HS is another memory-based algorithm inspired from a jazz
music orchestra when playing and improvising in their attempt
to achieve a progressively better harmony [96]. Landa-Torres et al.
[97] proposed a multi-objective HS algorithm for the efficient dis-
tribution of 24-h emergency units. This paradigm is essentially a
facility location problem that involves determining the optimum
locations of 24-h emergency resources and the optimum assign-
ment of patients to the resources through the existing medical care
infrastructure. Their algorithm redefines the grouping encoding to
reduce the dimension of the search space and employs the nondom-
inated sorting for solution evaluation. Simulations in a real scenario
based on geographic data of medical centers over the provinces of
Guadalajara and Cuenca, Spain, demonstrate the robustness and
applicability of the proposed approach.
Ozdamar and Yi [98] used greedy neighborhood search to
develop a fast constructive heuristic, named Path-Builder, to
support relief and evacuation operations. The method extends
the neighborhood structure to suit the problem’s special needs,
and selects partial paths to append to vehicle routes in the
operations, and thus exploits foreseeable opportunities within
the vehicles’ limited neighborhood. Considering the allocation
of scarce resources to repair a rural road network, Duque and
Sörensen [99] used GRASP and VNS to maximize the accessibility
of as many people as possible to the main cities or regional centers
where the economic and social infrastructure is usually located.
The efficiency of the approach is demonstrated by applying it to a
large real-life motivated instance.
Table 7 summarizes the characteristics of typical heuristic meth-
ods other than the EAs mentioned in the previous sections.
Table 7
The characteristics of other heuristic search approaches.
M-Pop Unc/Rand MOO PBO M-Pop
Jahangiri et al. [90]
Ma et al. [91]
Xie and Turnquist [94]
√ √
Zheng et al. [95]
√ √
Landa-Torres et al. [97]
Ozdamar and Yi [98]
√
Duque and Sörensen [99]
3.8. Hybrid algorithms
Different optimization methods have different design principles
and application areas. By exploiting the strengths of two or more
solution methods, we have a chance to obtain a powerful approach
that is much more competitive than any individual method. For
solving a large-scale multi-depots VRP in relief work, Peng et al.
[100] proposed a method combining GA with ACO and SA. Using a
feedback loop, the best result of GA is used to improve ant systems
and vice versa, in order to take advantage of both the fast conver-
gence of GA and the preciseness of ACO, and the SA procedure is
incorporated to avoid the local optimal. Zhang et al. [101] devel-
oped a hybrid AIS and ACO method for a route choice problem in
emergency logistics. The main idea is to obtain pheromone distribu-
tion from AIS, and optimize the solutions by basic ACO. Simulation
results show that the method has more effective timeliness than
basic ACO.
Hu et al. [102] studied the allocation of earthquake emergency
shelters which involves multiple sites and strict constraints, and
developed a new discrete PSO which uses a feasibility rule to handle
the constraints, and embeds a SA for potentially escaping from local
optima. The approach has been applied to the allocation of earth-
quake emergency shelters in the Zhuguang Block of Guangzhou
City, China.
Zhang and Fei [103] designed a novel approach that integrated
ACO with another SI method, fish swarm algorithm [104], to solve
a weak economy emergency logistics path optimization problem,
the aim of which is to gain a win–win result of the lowest cost
on condition of meeting all affected points time requirements. The
hybrid method, namely the fish swarm ant colony algorithm, takes
the advantage of the structure of the crowded degree factor in the
ant colony for improving the model of fish swarm in the ability of
global search. A similar research was conducted by Fei et al. [105],
where an ACO based on SA is proposed for path optimization in
emergency logistics.
For solving the problem of post-disaster infrastructure man-
agement, Kallioras et al. [106] proposed an integrated method,
where an improved HS algorithm is used for emergency inspection
scheduling, and an ACO algorithm is employed for solving the asso-
ciated routing problem. Comparative results show the proposed
method outperforms other algorithms including PSO, DE, the basic
HS, and the pure random search.
Becker et al. [107] introduce a new multi-agent algorithm for
search and rescue operations for exploration of unknown terrain.
The hybrid method uses the flood algorithm to a fast exploration
of the unknown terrain, and utilize the path optimizing features
of ACO for constructing short paths from points of interest back to
the base. The feasibility of the approach is demonstrated by agent-
based simulations.
Considering emergency railway wagon scheduling between
multiple source, center, and target stations, Zheng et al. [80] pre-
sented an integer programming model, the calculation of objectives
and constraints of which further include a set of nonlinear subprob-
lems. For efficiently tackling the problem, the authors proposed a
hybrid BBO algorithm which uses a local ring topology to avoid
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
premature convergence, employs the DE mutation to enhance
exploration, and takes some problem-specific mechanisms to fine-
tuning the search. Computational experiments and a real-world
case study demonstrate the effectiveness and the scalability of the
algorithm.
4. Experimental analysis on an integrated transportation
problem
In this section we test the performance of some typical EAs
on an emergency problem that integrates transportation plan-
ning depicted in Section 2.1 and vehicle routing in Section 2.3. The
problem extends model (1) with heterogeneous commodities and
vehicle capacity constraints as follows:

m 
n

K
min f = wjk t(xijk )
i=1 j=1 k=1

m
s.t. xijk ≥ bjk , j = 1, . . ., n, k = 1, . . ., K
i=1
n
xij ≤ aik , i = 1, . . ., m, k = 1, . . ., K
j=1
n 
 K
xijk ≥ ci , i = 1, . . ., m
j=1 k=1
xijk ≥ 0, i = 1, . . ., m, j = 1, . . ., n, k = 1, . . ., K
where K is the number of types of relief commodities, bjk and aik
are the demand and supply of commodity k of target j and source i
respectively, ci is the maximum vehicle capacity available at source
i, and wjk is the importance weight of demand item bjk . xijk , the
amount of commodity k from source i to target j, is the decision vari-
able, while t(xijk ) is the corresponding arrival time to be calculated
based on a vehicle routing subproblem for each source i. In real-
world relief operations, the integrated problem and its variations
occur much more frequently than the two individual subproblems
[52,108].
The evaluation of each solution to the above main problem
involves solving a VRP instance and thus is computationally expen-
sive. An algorithm with high search ability can effectively reduce
the number of function evaluations and improve the response
speed. Here we comparatively run the following six state-of-the-art
EAs on the main problem:
• The standard PSO 2007 [109] which equips the basic PSO with a
local random topology and a few fine-tuning adjustments.
• The self-adaptive DE (SaDE) [110] whose mutation schemes and
control parameters are self-adapted by learning from previous
generations.
• The Blended BBO (B-BBO) [111] which improves the basic BBO
with a blended migration operator.
• The GA combined with three fast improvement procedures [31].
• The GPSO [112] that combines PSO with a gradient-based local
search for accurate local exploration.
• The EBO [82] which enhances BBO with a local topology and the
combination of a global migration and a local migration.
In most real-world relief operations, the average number of tar-
gets assigned for each source is relatively small, and thus we employ
the branch-and-bound method [113] for solving VRP for all the EAs
for a fair comparison. We also use the following constraint handling
method in fitness comparison in all the EAs:
• A feasible solution is always fitter than an infeasible one.
561
• For two feasible solutions, the one with smaller objective function
value is fitter.
• For two infeasible solutions, the one with lower constraint viola-
tion is fitter.
The algorithms are tested on a set of 10 problem instances
from real-world relief operations occurred in China since 2008, the
dimensions of which range from 45 to 1080. The experiments are
conducted on a computer of Intel Core i5-2520M CPU and 8 GB
memory. For each instance, a maximum running time limit Tmax
(in minutes) is set for all the EAs according to the response time
requirement in the emergency, and each algorithm is run 30 times
with different random seeds. Table 8 summarizes the experimental
results.
The first two instances are respectively derived from the opera-
tions of two moderate earthquakes (the 2010 mag 4.6 earthquake
in Ali, Tibet autonomous region and the 2009 mag 4.7 earthquake
in Baicheng, Jilin province). The relief demands are not strong and
the delivery tasks are not heavy. As we can see, on instance #1
all the EAs always achieve the same result; on instance #2 four
EAs except SaDE and B-BBO always achieve the same result on #2.
(4) Post-hoc analysis shows that the results are real optimal solutions.
Note SaDE and B-BBO also frequently reach the optimum of #2. This
shows that there are no significant performance differences among
these popular EAs on the two small instances.
Instance #3 also represents a moderate earthquake relief oper-
ation (the 2011 mag 4.8 earthquake in Anqing, Anhui province),
but the earthquake occurred in a more developed region, where
the transportation network is much more complex, and thus it
is quite difficult to find the optimal relief delivery plan. Within
the limited running time, none of the algorithms reach the opti-
mum; EBO achieves the best result among the six EAs, but it has no
significant difference from GPSO, SaDE and SPO 2007.
Instance #4 represents the operation of the 2011 mag 5.8 earth-
quake in Yingjiang, Yunan province. Although the demand is not
strong, the operation involves much more supply centers, which
significantly enlarges the solution space and increases the num-
ber of local optima. On this instance, GPSO exhibits the best mean
performance and EBO ranks second, and the two methods show sig-
nificant performance improvement over the other four EAs, mainly
because their exploration abilities help to jump out of the local
optima. An interesting finding is that all the EAs have reached the
optimal solution among the 30 runs, which shows the random seed
also affects the algorithm performance to a certain degree on this
instance.
Instance #5, which represents the operation of the 2010 mag
6.0 earthquake in Chuxiong, Yunan province, is similar to #4 in
the huge number of local optima. Here EBO exhibits the best
mean performance and GPSO ranks second, showing significant
improvement over the other four. Except them, PSO 2007 can also
occasionally reach the optimum among the 30 runs.
For instance #6 (the 2010 Zhouqu mudslides), the relief demand
is strong and the solution space is large. However, the number of
local optimal is not so huge, mainly because the demand points are
located in a very small area. Within the limited time, GPSO and EBO
achieve much better results than the other EAs. However, it is found
that, when given enough running time (about 20–30 min), all the
EAs can reach very close to the same optimum, which is a unique
phenomenon among the test instances.
The last four instances are all large-scale operations with quite
large solution space and huge number of local optima. On #7
(the 2010 mag 7.3 earthquake in Yushu, Gansu province), GPSO
and EBO exhibit significant performance improvement over the
others. The mean result of EBO is the best, but it has no sig-
nificant difference from GPSO. While on the more difficult #8
(the 2013 mag 6.6 earthquake in Dingxi, Gansu), EBO exhibits
562 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566

Table 8
The experimental results of six EAs on the test problem instances (“mean” denotes the average result objective value over the 30 runs, “std” the corresponding standard
deviation, and “best” the best objective value among the 30 runs).

# Size (m × n × K) Tmax Metrics PSO2007 SaDE B-BBO GA GPSO EBO

1 3 × 5 ×3 1 Mean 1.58E+01 1.58E+01 1.58E+01 1.58E+01 1.58E+01 1.58E+01

Std 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00 0.00E+00
Best 1.58E+01 1.58E+01 1.58E+01 1.58E+01 1.58E+01 1.58E+01

2 8 × 3 ×2 1 Mean 5.09E+01 5.20E+01 5.27E+01 5.09E+01 5.09E+01 5.09E+01

Std 0.00E+00 2.11E+00 3.69E+00 0.00E+00 0.00E+00 0.00E+00
Best 5.09E+01 5.09E+01 5.09E+01 5.09E+01 5.09E+01 5.09E+01

3 8 × 5 ×3 2 Mean 3.78E+03 3.72E+03 3.96E+03 3.82E+03 3.64E+03 3.60E+03

Std 6.62E+02 3.94E+02 5.90E+02 7.79E+02 3.12E+03 2.98E+02
Best 3.48E+03 3.48E+03 3.57E+03 3.53E+03 3.48E+03 3.48E+03

4 12 × 4 ×3 2 Mean 7.05E+03 7.36E+03 7.78E+03 7.95E+03 6.80E+03 6.92E+03

Std 1.29E+03 1.20E+03 2.05E+03 2.52E+03 1.03E+03 9.98E+03
Best 6.55E+03 6.55E+03 6.55E+03 6.55E+03 6.55E+03 6.55E+03

5 6 × 6 ×5 2 Mean 6.15E+04 6.56E+04 7.90E+04 8.25E+04 6.00E+04 5.89E+04

Std 1.23E+04 9.78E+03 1.95E+04 2.23E+04 9.52E+04 8.90E+04
Best 5.30E+04 6.00E+04 6.00E+04 6.40E+04 5.30E+04 5.30E+04

6 16 × 3 ×4 2 Mean 9.50E+04 1.09E+05 1.72E+05 1.56E+05 8.29E+04 8.12E+04

Std 2.75E+04 2.87E+04 3.60E+04 3.09E+04 2.26E+04 1.97E+04
Best 8.01E+04 8.85E+04 1.49E+04 1.30E+04 7.59E+04 7.68E+04

7 21 × 3 ×7 3 Mean 7.58E+06 9.33E+06 2.74E+07 1.89E+07 5.75E+06 4.88E+06

Std 1.66E+06 1.86E+06 5.52E+06 4.72E+06 1.52E+06 1.16E+06
Best 6.30E+06 7.19E+06 2.35E+07 1.36E+07 4.25E+06 4.06E+06

8 15 × 5 ×6 3 Mean 3.37E+07 5.39E+07 1.76E+08 1.28E+08 2.10E+07 1.55E+07

Std 6.63E+06 9.60E+06 3.49E+07 1.76E+07 5.21E+06 3.22E+06
Best 2.58E+07 4.24E+07 1.31E+08 1.08E+08 1.65E+07 1.28E+07

9 18 × 10 × 5 5 Mean 1.84E+10 3.21E+10 1.75E+11 2.05E+11 7.69E+09 3.56E+09

Std 2.64E+09 5.61E+09 2.73E+10 3.90E+10 2.53E+09 7.99E+08
Best 1.67E+10 2.67E+10 1.49E+11 1.70E+11 5.06E+09 2.76E+09

10 20 × 9 ×6 5 Mean 8.92E+11 1.62E+12 1.73E+13 5.85E+13 1.49E+11 1.05E+11

Std 1.50E+11 3.89E+11 2.78E+12 1.33E+12 4.08E+10 2.36E+10
Best 7.76E+11 1.27E+12 1.33E+13 4.26E+13 1.16E+11 9.39E+10

significant performance advantage over all the other EAs including
GPSO.
Instance #9 represents the operation of the 2008 mag 8.0
earthquake in Wenchuan, Sichuang province – the most serious
disaster in China in the recent decade. Among all the EAs, EBO
also exhibits the best performance. Nevertheless, according to the
decision-maker/experts of the relief operation, even the mean
result of EBO is inefficient; in fact, only the best result of EBO
(reached thrice among the 30 runs of the algorithm) is considered
efficient.
Instance #10, the operation of the 2013 mag 7.0 earthquake in
Ya’an, Sichuang province, is even more difficult than #9 because
the disaster response system in China has been much improved
since 2008, and thus the event involves much more relief demands
and forces. On this instance, in general, none of EAs can reach a
solution considered efficient according to post-hoc analysis. The
best solution of EBO (reached twice among the 30 runs) is accept-
able, but not so satisfactory to the decision-maker. In fact, we
can obtain an efficient solution (with an objective value around
3.50E+10) after about 60 EBO runs, each with 1 hour running
time.
In summary, EBO and GPSO generally perform better the other
four EAs on the test instances, and their performance advantages
increase with the problem size/difficulty. On relatively large-size
instances #7–#10, the performance of B-BBO and GA becomes
much inferior, because their global search abilities are not strong
and thus have difficulty in exploring a very large solution space.
PSO 2007 has a fast convergence speed, but its results are not
very satisfactory on large-size instances due to the limited local
exploitation ability. SaDE, an EA that is competitive on a wide
range of benchmark problems, exhibits mediocre performance
on our test instances, mainly because it requires feedback from
previous generations to adjust the search mechanism, but the max-
imum running time is too limited to allow enough feedback. In
comparison, EBO and GPSO balance exploration and exploitation
much more effectively and thus exhibit much better performance.
However, on extremely large instances such as #10, the required
running time (or number of generations) for producing quality solu-
tions is contradictory to the limited response time, which is an open
problem faced by current research.
From the experiments on the typical problem, we can draw the
following lessons:
• Generally the balance between exploration and exploitation is
a key to achieve promising performance. Particularly, under
emergency conditions, exploration ability (such as provide by
PSO and DE) can be more important in producing acceptable
solutions within limited response time, while the design of effec-
tive yet efficient exploitation mechanisms can be much more
difficult.
• Self-adaptive mechanisms, which are useful for a wide vari-
ety of problems, may be not effective for emergency problems,
because the response time often limits the number of algorithm
generations and thus cannot allow enough feedback to enhance
performance.
• For different types of optimization problems in disaster relief
operations, there is no a magic bullet algorithm that can be uni-
formly better than any other algorithm (known as the “No Free
Lunch Theorem” [114]). Thereby, it will be very helpful to obtain
a priori knowledge about the problem in disaster preparedness,
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566 563

and utilize them in the design of problem-specific mechanisms to
facility the search for quality solutions, especially for very large-
scale operations.
5. Strengths, limitations and future perspectives
From the survey of recent literature, we can clearly see that
EAs represents highly competitive alternatives to conventional
approaches for complex disaster relief operational problems where
these latter methods are often unaffordable or require unaccept-
able assumptions. The main strengths of EAs for these critical
applications can be summarized as follows:
• EAs are general metaheuristics that can be easily applied or
adapted to a large variety of operational problems, in particular
those new problems with little prior knowledge.
• Empirically, EAs are capable of finding high quality solutions
for most problems in consideration, although theoretically their
convergence to global optima is only guaranteed in a weak prob-
abilistic sense.
• Population-based search mechanism and elaborately designed
evolutionary operators advance solutions in a quite efficient
manner, resulting in considerable reduction in computational
resource and thus meeting stringent response time requirements
for problem solving.
• EAs are particularly effective in handing problems with various
and often incompatible types of objectives and/or constraints,
which are very common in emergency problems under stressful
disaster conditions.
• EAs perform well on problems with noise and uncertain param-
eters, and produce solutions robust and adaptable to changing
environments.
• The optimization processes of EAs are flexible, and can be tuned
to a great extent to the problems at hand.
algorithm [118], etc., have been proposed and demonstrated their
effectiveness and efficiency on a variety of problems. There is
plenty of room for these new metaheuristics as well as their
hybridization with existing methods in the area of disaster
relief.
• Pareto-based multiobjective optimization and decision support.
MOEA embodies an exciting opportunity to realize the full poten-
tial of the EA concept [119], and it is expected that studies on the
optimization of disaster relief operations would benefit greatly
from the exploitation of the power of MOEA techniques. On the
other hand, multi-criteria decision making techniques including
preference articulation, TOPSIS [120] and Pareto-front visualiza-
tion, should be effectively utilized to help decision makers to
understand the resulting Pareto-optimal solutions and choose the
most appropriate one(s) for implementation.
• Hyper-heuristics. Defined as “heuristics to choose heuristics”,
hyper-heuristics [121] are high level-heuristics that perform
search over a set of low-level heuristics, and can often achieve not
only better solutions but also much higher flexibility and robust-
ness on very difficult problems. Hyper-heuristics have been
successfully applied to a wide range of planning and scheduling
problems, and their potential in emergency optimization prob-
lems remains to be explored.
• Test suites and real-world applications. It is of vital importance
to determine to which degree the solutions of EAs would meet
the stringent quality and safety requirements of relief oper-
ations. However, there are few benchmark problems publicly
available for this purpose. It is desirable for the community
to provide more standard test suites, test case generators, as
well as evaluation tools for assessment and analysis of dif-
ferent EAs. More importantly, proposed solution approaches
are required to be fully implemented and verified on cho-
sen real-world problems to demonstrate their practicality and
effectiveness.

6. Conclusion
However, the application of EAs to disaster relief operational
problems, in comparison with other common operational prob-
The paper gives a survey of research advances in evolu-
lems, still has many limitations and disadvantages. First, GAs are
tionary optimization applied to disaster relief operations. The
in the majority of EAs used for emergency operational problems;
survey is not intended to be all embracing. The main purpose
Many state-of-the-art methods (in particular those hybrid algo-
is to help readers have a general overview of the state-of-the-
rithms) which have been demonstrated to be much more efficient
arts and easily refer suitable methods in practical applications,
on benchmark problems, have been less explored.
given that the works described here are indicative of the breadth
Second, although MOEAs have been proved to be singularly
of emergency operational problems where EAs have proved
successful in effectively addressing MOO problems, most research
profitable.
on MOO in emergency operations still simply combines multiple
As a clear finding of the survey, EAs have found a growing num-
objectives into a single one, the results of which are much less
ber of applications in disaster relief operations in recent years.
comprehensive to support decision makers who have to resolve
However, in comparison with other typical areas in OR, the poten-
a divergence of interest.
tial of evolutionary computing in emergency operational problems
Third, most computational experiments are conducted on test
is far from being realized. The true challenge that we must over-
problems generated randomly. Only a few research uses real-world
come, and this is true of soft computing in general, is to convince
data for testing. More importantly, reports on successful applica-
sceptical decision makers that the new methods are capable of pro-
tions of EAs to real-world relief operations are very scarce. Only in
ducing results worthy of their trust. To achieve this, the methods
very recent years, several works on bio-inspired methods for trans-
have to be actually applied to and demonstrate their performance
portation planning in the 2011 Yingjiang earthquake [52] and the
on more real-world operations.
2013 Lushan earthquake [80] and repair scheduling in the 2013 Dixi
earthquake [81] have been reported.
The future use of evolutionary computation in disaster relief Acknowledgements
operations is tied to both the increasing difficulty of practical prob-
lems and the advances of EAs themselves. In view of the current This work was supported by Project 61020106009, 61105073,
limitations in this area, the following aspects deserve attention in 613250211 and 61473263 supported by Natural Science Founda-
future development initiatives: tion of China.

• Emerging metaheuristic optimization methods. In recent years, a Appendix A.

number of new EAs, including BBO, cuckoo search [115], paddy
field algorithm [116], bat-inspired algorithm [117], fireworks See Table 9.
564 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566

Table 9
The summary of the research works classified by problem category.

General TP Allocation &location Routing Repair Integrated

GA Tuson et al. [26] Wang and Zhang [32] Ma et al. [39] Sato and Ichii [48] Chang et al. [51]
Feng and Wen [27] Kongsomsaksakul et al. [33] Lin et al. [40] Chen and Tzeng [49] Zheng and Ling [52]
Na and Zhi [28] Jia et al. [34] Zhang et al. [41] Zhang and Lu [50]
Nolz et al. [29] Han and Zhang [35] Hamedi et al. [42]
Yang et al. [30] Chai et al. [36] Chung et al. [43]
Berkoune et al. [31] Tzeng and Chen [37] Liu et al. [44]
Yang et al. [38] Song et al. [45]
Saadatseresht et al. [46]
Ma et al. [47]

ES Xie and Zhang [59] Barrachina et al. [60]

EP Wang and Jin [61]

PSO Pang et al. [66] Uno et al. [64] Wang et al. [70] Gan et al. [68]
Bozorgi-Amiri et al. [67] Chai et al. [65] Zheng et al. [71] Tian et al. [69]
Hu et al. [102]

ACO Lim et al. [74] Yuan and Wang [75] Yi and Kumar [76]
Yan and Shih [77]

AIS Hu [85] Liu and Zhang [87]

Zhang et al. [88]

BBO Zheng et al. [82] Zheng et al. [81] Zheng et al. [80]

SA Ma et al. [91] Jahangiri et al. [90]

TS Xie and Turnquist [94] Zheng et al. [95]

HS Landa-Torres et al. [97]

VNS Ozdamar and Yi [98] Duque and Söorensen [99]

Hybrid Peng et al. [100] Kallioras et al. [106]

Zhang et al. [101] Becker et al. [107]
Zhang and Fei [103] Zheng et al. [80]
Fei et al. [105]

References
[1] M.R. Garey, D.S. Johnson, Computers and Intractability: A Guide to the Theory
of NP-Completeness, W.H. Freeman, San Francisco, 1979.
[2] S. Yang, Y.-S. Ong, Y. Jin, Evolutionary Computation in Dynamic and Uncertain
Environments, Studies in Computational Intelligence, Springer-Verlag New
York, Inc., Secaucus, NJ, USA, 2007.
[3] W. Pedrycz, Fuzzy evolutionary computing, Soft Comput. 2 (1998) 61–72.
[4] C. Grosan, A. Abraham, H. Ishibuchi, Hybrid Evolutionary Algorithms,
Springer, Berlin, 2007.
[5] J.H. Holland, Adaptation in Natural and Artificial Systems: An Introductory
Analysis with Applications to Biology, Control and Artificial Intelligence, MIT
Press, Cambridge, MA, USA, 1975.
[6] J. Kennedy, R. Eberhart, Particle swarm optimization, in: IEEE Interna-
tional Conference on Neural Networks, vol. 4, 1995, pp. 1942–1948,
http://dx.doi.org/10.1109/ICNN.1995.488968.
[7] M. Dorigo, G. Di Caro, Ant colony optimization: a new meta-heuristic, in:
Proceedings of the 1999 Congress on Evolutionary Computation, IEEE, vol. 2,
1999, pp. 1470–1477.
[8] W. Fan, The consideration and suggestion of scientific problems on emergency
management of national sudden public events, Sci. Found. China 21 (2007)
71–76.
[9] X. Zuo, Q. Ran, W. Gu, Research on emergency logistics: situations
and development trends, in: IEEE International Conference on Grey Sys-
tems and Intelligent Services, 2009, pp. 1638–1641, http://dx.doi.org/
10.1109/GSIS.2009.5408177.
[10] V. Srinivasan, G.L. Thompson, Benefit-cost analysis of coding techniques for
the primal transportation algorithm, J. ACM 20 (1973) 194–213.
[11] G.A. Vignaux, Z. Michalewicz, A genetic algorithm for the linear transportation
problem, IEEE Trans. Syst. Man Cybern. 21 (1991) 445–452.
[12] H. Lau, T. Chan, W. Tsui, F. Chan, G. Ho, K. Choy, A fuzzy guided multi-objective
evolutionary algorithm model for solving transportation problem, Expert
Syst. Appl. 36 (2009) 8255–8268.
[13] L. Özdamar, E. Ekinci, B. Küçükyazici, Emergency logistics planning in natural
disasters, Ann. Oper. Res. 129 (2004) 217–245.
[14] M. Melo, S. Nickel, F.S. da Gama, Facility location and supply chain manage-
ment - a review, Eur. J. Oper. Res. 196 (2009) 401–412.
[15] A.B. Arabani, R.Z. Farahani, Facility location dynamics: an overview of classi-
fications and applications, Comput. Ind. Eng. 62 (2012) 408–420.
[16] R. Church, C.R. VELLE, The maximal covering location problem, Pap. Reg. Sci.
32 (1974) 101–118.
[17] K. Hogan, C. Revelle, Concepts and applications of backup coverage, Manag.
Sci. 32 (1986) 1434–1444.
[18] B. Eksioglu, A.V. Vural, A. Reisman, The vehicle routing problem: a taxonomic
review, Comput. Ind. Eng. 57 (2009) 1472–1483.
[19] M. Pinedo, Scheduling Theory, Algorithms, and Systems, 2nd ed., Prentice Hall,
New York, 2002.
[20] S. Yan, Y.-L. Shih, A time-space network model for work team scheduling after
a major disaster, J. Chin. Inst. Eng. 30 (2007) 63–75.
[21] S. Yan, Y.-L. Shih, Optimal scheduling of emergency roadway repair and sub-
sequent relief distribution, Comput. Oper. Res. 36 (2009) 2049–2065.
[22] K. Miettinen, Nonlinear Multiobjective Optimization, Kluwer, Boston, 1999.
[23] K. Deb, Multi-objective optimization using evolutionary algorithms, John
Wiley & Sons, Chichester, 2001.
[24] C.A.C. Coello, D.A. Van Veldhuizen, G.B. Lamont, Evolutionary Algorithms for
Solving Multi-Objective Problems, Springer, New York, 2002.
[25] Q. Zhao, A general framework for cooperative co-evolutionary algo-
rithms: a society model, in: Evolutionary Computation Proceedings,
IEEE World Congress on Computational Intelligence, 1998, pp. 57–62,
http://dx.doi.org/10.1109/ICEC.1998.699134.
[26] A. Tuson, R. Wheeler, P. Ross, An evolutionary/meta-heuristic approach to
emergency resource redistribution in the developing world, in: Artificial
Neural Nets and Genetic Algorithms, Springer Vienna, 1998, pp. 329–332,
http://dx.doi.org/10.1007/978-3-7091-6492-1 72.
[27] C.-M. Feng, C.-C. Wen, Traffic control management for earthquake-raided
area, J. East. Asia Soc. Transp. Stud. 5 (2003) 3261–3275.
[28] L. Na, L. Zhi, Emergency relief goods multi-mode transportation based
on genetic algorithm, in: Second International Conference on Intelligent
Computation Technology and Automation, vol. 1, 2009, pp. 181–184,
http://dx.doi.org/10.1109/ICICTA.2009.52.
[29] P.C. Nolz, K.F. Doerner, W.J. Gutjahr, R.F. Hartl, A bi-objective metaheuristic for
disaster relief operation planning, in: C. Coello Coello, C. Dhaenens, L. Jourdan
(Eds.), Advances in Multi-Objective Nature Inspired Computing, vol. 272 of
Studies in Computational Intelligence, Springer Berlin Heidelberg, 2010, pp.
167–187, http://dx.doi.org/10.1007/978-3-642-11218-8 8.
[30] L. Yang, Z. Gao, K. Li, Railway freight transportation planning with mixed
uncertainty of randomness and fuzziness, Appl. Soft Comput. 11 (2011)
778–792.
[31] D. Berkoune, J. Renaud, M. Rekik, A. Ruiz, Transportation in disaster response
operations, Socio-Econ. Plan. Sci. 46 (2012) 23–32.
[32] D.-W. Wang, G.-X. Zhang, Model and algorithm to optimize location of cata-
strophic rescue center, J. Northeast. Univ. (Nat. Sci.) 26 (2005) 953–956.
 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
[33] S. Kongsomsaksakul, C. Yang, A. Chen, Shelter location-allocation model for
flood evacuation planning, J. East. Asia Soc. Transp. Stud. 6 (2005) 4237–
4252.
[34] H. Jia, F. Ordóǹez, M.M. Dessouky, Solution approaches for facility location
of medical supplies for large-scale emergencies, Comput. Ind. Eng. 52 (2007)
257–276.
[35] C.-f. Han, C. Zhang, Genetic algorithm for solving problems in emergency man-
agement, in: Fifth International Conference on Natural Computation, vol. 4,
2009, pp. 259–264, http://dx.doi.org/10.1109/ICNC.2009.333.
[36] G. Chai, C. Fang, X. Gao, Q. Zhao, A cost-based study on highway traffic
emergency rescue sites location using heuristic genetic algorithm, J. Comput.
Inform. Syst. 7 (2011) 507–514.
[37] G.-H. Tzeng, Y.-W. Chen, The optimal location of airport fire stations: a
fuzzy multi-objective programming and revised genetic algorithm approach,
Transp. Plan. Technol. 23 (1999) 37–55.
[38] L. Yang, B.F. Jones, S.-H. Yang, A fuzzy multi-objective programming for opti-
mization of fire station locations through genetic algorithms, Eur. J. Oper. Res.
181 (2007) 903–915.
[39] Y. Ma, Y. Tian, W. Hou, Dynamic path optimization of emergency trans-
port based on hierarchical genetic algorithm, in: Y. Tan, Y. Shi, K.
Tan (Eds.), Advances in Swarm Intelligence, vol. 6145 of Lecture Notes
in Computer Science, Springer Berlin Heidelberg, 2010, pp. 445–452,
http://dx.doi.org/10.1007/978-3-642-13495-1 55.
[40] Y.-H. Lin, R. Batta, P.A. Rogerson, A. Blatt, M. Flanigan, A logistics model for
emergency supply of critical items in the aftermath of a disaster, Socio-Econ.
Plan. Sci. 45 (2011) 132–145.
[41] J. Zhang, J. Peng, Z. Xu, W. Xu, SDVRP model for emergency logistics
and evolutionary heuristic approach, in: International Conference on Auto-
matic Control and Artificial Intelligence (ACAI 2012), 2012, pp. 1809–1812,
http://dx.doi.org/10.1049/cp.2012.1342.
[42] M. Hamedi, A. Haghani, S. Yang, Reliable transportation of humanitarian sup-
plies in disaster response: model and heuristic, Proc. – Social Behav. Sci. 54
(2012) 1205–1219.
[43] L.-Y. Chung, T.-Y. Chou, C.-C. Lee, Multiobjective dynamic length genetic
algorithm to solve the emergency logistic problem, in: 2012 IEEE/ASME
International Conference on Advanced Intelligent Mechatronics, 2012, pp.
1147–1152, http://dx.doi.org/10.1109/AIM.2012.6266026.
[44] Y. Liu, X. Lai, G. Chang, Two-level integrated optimization system for planning
of emergency evacuation, J. Transp. Eng. 132 (2006) 800–807.
[45] R. Song, S. He, L. Zhang, Optimum transit operations during the emergency
evacuations, J. Transp. Syst. Eng. Inform. Technol. 9 (2009) 154–160.
[46] M. Saadatseresht, A. Mansourian, M. Taleai, Evacuation planning using mul-
tiobjective evolutionary optimization approach, Eur. J. Oper. Res. 198 (2009)
305–314.
[47] C. Ma, Y. Li, R. He, F. Wu, L. Sun, B. Qi, Multi-objective fuzzy optimisation
method for emergency evacuation based on hybrid intelligence algorithm
under fuzzy environment, Int. J. Comput. Appl. Technol. 46 (2013) 228–
235.
[48] T. Sato, K. Ichii, Optimization of post-earthquake restoration of lifeline
networks using genetic algorithms, Jpn. Soc. Civil Eng. 53 (1996) 245–256.
[49] Y.-W. Chen, G.-H. Tzeng, A fuzzy multi-objective model for reconstructing the
post-quake road-network by genetic algorithm, Int. J. Fuzzy Syst. 1 (1999)
85–95.
[50] T.Z. Zhang, Y.M. Lu, Study on simulation and optimization of the road rush-
repair model after disaster, Appl. Mech. Mater. 50 (2011) 298–303.
[51] F.-S. Chang, J.-S. Wu, C.-N. Lee, H.-C. Shen, Greedy-search-based multi-
objective genetic algorithm for emergency logistics scheduling, Expert Syst.
Appl. 41 (2014) 2947–2956.
[52] Y.-J. Zheng, H.-F. Ling, Emergency transportation planning in disaster relief
supply chain management: a cooperative fuzzy optimization approach, Soft
Comput. 17 (2013) 1301–1314.
[53] G.F. Mott, Optimising flowshop scheduling through adaptive genetic algo-
rithms, Chemistry Part II Thesis, Ph.D. thesis, Oxford University, Oxford, 1990.
[54] K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A fast and elitist multiobjective
genetic algorithm: NSGA-II, IEEE Trans. Evol. Comput. 6 (2002) 182–197.
[55] B. Liu, Uncertainty Theory, 2nd ed., Springer-Verlag, Berlin, 2007.
[56] I. Rechenberg, Evolutions strategies: Optimiering technischer systeme nach
prinzipien der biologischen evolution, Frommberg-Holzboog, Stuttgart, 1973.
[57] L.J. Fogel, A.J. Owens, M.J. Walsh, Artificial Intelligence Through Simulated
Evolution, John Wiley & Sons, Chichester, 1967.
[58] H.-G. Beyer, H.-P. Schwefel, Evolution strategies – a comprehensive introduc-
tion, Nat. Comput. 1 (2002) 3–52.
[59] H. Xie, L. Zhang, Using dual-gene variation hybrid evolution strategy to
construct fire emergency decision-making model, in: 2nd International Con-
ference on Advanced Computer Control (ICACC), 2010, vol. 5, 2010, pp.
103–106, http://dx.doi.org/10.1109/ICACC.2010.5487295.
[60] J. Barrachina, P. Garrido, M. Fogue, F.J. Martinez, J.-C. Cano, C.T. Calafate,
P. Manzoni, Reducing emergency services arrival time by using vehicu-
lar communications and evolution strategies, Expert Syst. Appl. 41 (2014)
1206–1217.
[61] Y. Wang, F. Jin, Research on improved evolutionary programming algorithm
for solving the problem of emergency resources scheduling, Oper. Res. Manag.
Sci. 21 (2012) 29–33.
[62] R. Storn, K. Price, Differential evolution – a simple and efficient heuristic
for global optimization over continuous spaces, J. Global Optim. 11 (1997)
341–359.
565
[63] R. Eberhart, Y. Shi, Tracking and optimizing dynamic systems with particle
swarms, in: Proceedings of the 2001 Congress on Evolutionary Computation,
vol. 1, 2001, pp. 94–100, http://dx.doi.org/10.1109/CEC.2001.934376.
[64] T. Uno, K. Kato, H. Katagiri, An application of interactive fuzzy satisficing
approach with particle swarm optimization for multiobjective emergency
facility location problem with a-distance, in: IEEE Symposium on Compu-
tational Intelligence in Multicriteria Decision Making, 2007, pp. 368–373,
http://dx.doi.org/10.1109/MCDM.2007.369115.
[65] G. Chai, Y.-Y. Sun, C.-H. Zhu, Particle swarm optimization algorithm
for emergency resource allocation on expressway, in: Proceedings of
the First ACM/SIGEVO Summit on Genetic and Evolutionary Compu-
tation, GEC 2009, ACM, New York, NY, USA, 2009, pp. 135–140,
http://dx.doi.org/10.1145/1543834.1543854.
[66] H. Pang, N. Liu, Q. Wu, Decision-making model for transportation and distri-
bution of emergency materials and its modified particle swarm optimization
algorithm, Cont. Decis. 27 (2011) 871–874.
[67] A. Bozorgi-Amiri, M. Jabalameli, M. Alinaghian, M. Heydari, A modified
particle swarm optimization for disaster relief logistics under uncertain envi-
ronment, Int. J. Adv. Manuf. Technol. 60 (2012) 357–371.
[68] X. Gan, Y. Wang, Y. Yu, B. Niu, An emergency vehicle scheduling problem with
time utility based on particle swarm optimization, in: D.-S. Huang, K.-H. Jo,
Y.-Q. Zhou, K. Han (Eds.), Intelligent Computing Theories and Technology, vol.
7996 of Lecture Notes in Computer Science, Springer Berlin Heidelberg, 2013,
pp. 614–623, http://dx.doi.org/10.1007/978-3-642-39482-9 71.
[69] J. Tian, W.-Z. Ma, Y.-L. Wang, Emergency supplies distributing and vehicle
routes programming based on particle swarm optimization, Syst. Eng. – The-
ory Pract. 31 (2011) 898–906.
[70] C. Wang, B. Yang, L. Li, H. Huang, A modified particle swarm optimization-
based human behavior modeling for emergency evacuation simulation
system, in: International Conference on Information and Automation, 2008.
ICIA 2008, 2008, pp. 23–28, http://dx.doi.org/10.1109/ICINFA.2008.4607962.
[71] Y. Zheng, H. Ling, J. Xue, S. Chen, Population classification in fire evacuation: A
multiobjective particle swarm optimization approach, IEEE Trans. Evol. Com-
put. 18 (2014) 70–81.
[72] J.M. Mulvey, R.J. Vanderbei, S.A. Zenios, Robust optimization of large-scale
systems, Oper. Res. 43 (1995) 264–281.
[73] Q. Zhu, J. Hu, W. Cai, L. Henschen, A new robot navigation algorithm for
dynamic unknown environments based on dynamic path re-computation and
an improved scout ant algorithm, Appl. Soft Comput. 11 (2011) 4667–4676.
[74] K.K. Lim, Y.-S. Ong, M.H. Lim, X. Chen, A. Agarwal, Hybrid ant colony algo-
rithms for path planning in sparse graphs, Soft Comput. 12 (2008) 981–994.
[75] Y. Yuan, D. Wang, Path selection model and algorithm for emergency logistics
management, Comput. Ind. Eng. 56 (2009) 1081–1094.
[76] W. Yi, A. Kumar, Ant colony optimization for disaster relief operations, Transp.
Res. E: Logist. Transp. Rev. 43 (2007) 660–672.
[77] S. Yan, Y.-L. Shih, An ant colony system-based hybrid algorithm for an emer-
gency roadway repair time-space network flow problem, Transportmetrica 8
(2012) 361–386.
[78] D. Simon, Biogeography-based optimization, IEEE Trans. Evol. Comput. 12
(2008) 702–713.
[79] K.P. Iyer, V. Shanthi, Location based emergency response management system
through satellites in road networks, UACEE Int. J. Comput. Sci. Appl. 3 (2012)
1–5.
[80] Y.-J. Zheng, H.-F. Ling, H.-H. Shi, H.-S. Chen, S.-Y. Chen, Emergency rail-
way wagon scheduling by hybrid biogeography-based optimization, Comput.
Oper. Res. 43 (2014) 1–8.
[81] Y.-J. Zheng, H.-F. Ling, X.-L. Xu, S.-Y. Chen, Emergency engineering rescue
scheduling and its application in disaster relief operations in China, Int. Trans.
Oper. Res. (2014) (in press).
[82] Y.-J. Zheng, H.-F. Ling, J.-Y. Xue, Ecogeography-based optimization: enhancing
biogeography-based optimization with ecogeographic barriers and differen-
tiations, Comput. Oper. Res. 50 (2014) 115–127.
[83] J.E. Hunt, D.E. Cooke, Learning using an artificial immune system, J. Netw.
Comput. Appl. 19 (1996) 189–212.
[84] Z.-H. Hu, A network for emergency logistics management inspired by immune
multi-affinity model, in: International Conference on Information Manage-
ment, Innovation Management and Industrial Engineering, 2009, vol. 4, 2009,
pp. 22–25, http://dx.doi.org/10.1109/ICIII.2009.467.
[85] Z.-H. Hu, A container multimodal transportation scheduling approach based
on immune affinity model for emergency relief, Expert Syst. Appl. 38 (2011)
2632–2639.
[86] X. Chen, G.-z. Tan, B. Jiang, Real-time optimal path planning for mobile robots
based on immune genetic algorithm, J. Cent. South Univ. (Sci. Technol.) 39
(2008) 577–583.
[87] L.-F. Liu, S.-Q. Zhang, Dual aircrafts path planning for forest-fire rescue by
artificial immune algorithm, Adv. Mater. Res. 1659 (2012) 532–533.
[88] X. Zhang, Z. Zhang, Y. Zhang, D. Wei, Y. Deng, Route selection for emergency
logistics management: a bio-inspired algorithm, Safety Sci. 54 (2013) 87–91.
[89] S.D.G. Kirkpatrick Jr., M.P. Vecchi, Optimization by simulated annealing, Sci-
ence 220 (1983) 671–680.
[90] A. Jahangiri, S. Afandizadeh, N. Kalantari, The optimization of traffic signal
timing for emergency evacuation using the simulated annealing algorithm,
Transport 26 (2011) 133–140.
[91] C. Ma, Y. Li, R. He, B. Qi, A. Diao, Research on location problem of emergency
service facilities based on genetic-simulated annealing algorithm, Int. J. Wirel.
Mobile Comput. 5 (2012) 206–211.
566 Y.-J. Zheng et al. / Applied Soft Computing 27 (2015) 553–566
[92] F. Glover, Tabu search – part I, ORSA J. Comput. 1 (1989) 190–206.
[93] F. Glover, Tabu search – part II, ORSA J. Comput. 2 (1990) 4–32.
[94] C. Xie, M.A. Turnquist, Lane-based evacuation network optimization: an inte-
grated Lagrangian relaxation and tabu search approach, Transp. Res. C: Emerg.
Technol. 19 (2011) 40–63.
[95] Y. Zheng, S. Chen, H. Ling, Efficient multi-objective tabu search for emergency
equipment maintenance scheduling in disaster rescue, Optim. Lett. 7 (2013)
89–100.
[96] Z.W. Geem, J.H. Kim, G. Loganathan, A new heuristic optimization algorithm:
harmony search, Simulation 76 (2001) 60–68.
[97] I. Landa-Torres, D. Manjarres, S. Salcedo-Sanz, J.D. Ser, S. Gil-Lopez, A multi-
objective grouping harmony search algorithm for the optimal distribution of
24-hour medical emergency units, Expert Syst. Appl. 40 (2013) 2343–2349.
[98] L. Ozdamar, W. Yi, Greedy neighborhood search for disaster relief and evacu-
ation logistics, IEEE Intell. Syst. 23 (2008) 14–23.
[99] P.M. Duque, K. Sörensen, A grasp metaheuristic to improve accessibility after
a disaster, OR Spectr. 33 (2011) 525–542.
[100] J. Peng, W. Xu, J. Yang, A hybrid heuristic algorithm for large scale
emergency logistics, in: Second International Conference on Intelligent
Computation Technology and Automation, vol. 3, 2009, pp. 899–902,
http://dx.doi.org/10.1109/ICICTA.2009.682.
[101] L.-Y. Zhang, T. Fei, X. Zhang, Y. Li, J. Yi, Research about immune ant colony opti-
mization in emergency logistics transportation route choice, in: M. Zhao, J. Sha
(Eds.), Communications and Information Processing, volume 288 of Commu-
nications in Computer and Information Science, Springer Berlin Heidelberg,
2012, pp. 430–437, http://dx.doi.org/10.1007/978-3-642-31965-5 50.
[102] F. Hu, W. Xu, X. Li, A modified particle swarm optimization algorithm for
optimal allocation of earthquake emergency shelters, Int. J. Geogr. Inform.
Sci. 26 (2012) 1643–1666.
[103] J. Zhang, T. Fei, The weak economy emergency logistics path optimiza-
tion algorithm based on fish swarm ant colony algorithm, in: J. Lei, F.
Wang, H. Deng, D. Miao (Eds.), Emerging Research in Artificial Intel-
ligence and Computational Intelligence, Communications in Computer
and Information Science, Springer Berlin Heidelberg, 2012, pp. 350–356,
http://dx.doi.org/10.1007/978-3-642-34240-0 46.
[104] X.-L. Li, Z.-J. Shao, J.-X. Qian, An optimizing method based on autonomous
animats: fish–swarm algorithm, Syst. Eng. Theory Pract. 11 (2002) 32–38.
[105] T. Fei, L.-y. Zhang, J. Zhang, Y. Li, X. Liu, C. Zhu, Research of emergency
logistics distribution routing optimization based on simulated annealing
ant colony optimization, in: J. Lei, F. Wang, H. Deng, D. Miao (Eds.), Arti-
ficial Intelligence and Computational Intelligence, vol. 7530 of Lecture
Notes in Computer Science, Springer Berlin Heidelberg, 2012, pp. 177–183,
http://dx.doi.org/10.1007/978-3-642-33478-8 23.
[106] N.A. Kallioras, N.D. Lagaros, M.G. Karlaftis, An improved harmony search
algorithm for emergency inspection scheduling, Eng. Optim. (2013),
http://dx.doi.org/10.1080/0305215X.2013.854351 (online).
[107] M. Becker, F. Blatt, H. Szczerbicka, A multi-agent flooding algorithm for
search and rescue operations in unknown terrain, in: M. Klusch, M. Thimm,
M. Paprzycki (Eds.), Multiagent System Technologies, vol. 8076 of Lecture
Notes in Computer Science, Springer Berlin Heidelberg, 2013, pp. 19–28,
http://dx.doi.org/10.1007/978-3-642-40776-5 5.
[108] Y.-J. Zheng, S.-Y. Chen, Cooperative particle swarm optimization for multiob-
jective transportation planning, Appl. Intell. 39 (2013) 202–216.
[109] M. Clerc, Standard PSO 2007, 2010 http://www.particleswarm.info
[110] A.K. Qin, V.L. Huang, P. Suganthan, Differential evolution algorithm with strat-
egy adaptation for global numerical optimization, IEEE Trans. Evol. Comput.
13 (2009) 398–417.
[111] H. Ma, D. Simon, Blended biogeography-based optimization for constrained
optimization, Eng. Appl. Artif. Intell. 24 (2011) 517–525.
[112] M.M. Noel, A new gradient based particle swarm optimization algorithm
for accurate computation of global minimum, Appl. Soft Comput. 12 (2012)
353–359.
[113] M. Fischetti, P. Toth, D. Vigo, A branch-and-bound algorithm for the capac-
itated vehicle routing problem on directed graphs, Oper. Res. 42 (1994)
846–859.
[114] D. Wolpert, W. Macready, No free lunch theorems for optimization, IEEE
Trans. Evol. Comput. 1 (1997) 67–82.
[115] X.-S. Yang, S. Deb, Cuckoo search via lévy flights, in: World Congress on Nature
Biologically Inspired Computing, 2009, pp. 210–214, http://dx.doi.org/10.
1109/NABIC.2009.5393690.
[116] U. Premaratne, J. Samarabandu, T. Sidhu, A new biologically inspired
optimization algorithm, in: International Conference on Industrial
and Information Systems, 2009, pp. 279–284, http://dx.doi.org/10.
1109/ICIINFS.2009.5429852.
[117] X.-S. Yang, A new metaheuristic bat-inspired algorithm, in: J.R. González, D.A.
Pelta, C. Cruz, G. Terrazas, N. Krasnogor (Eds.), Nature Inspired Cooperative
Strategies for Optimization, vol. 284 of Studies in Computational Intelligence,
Springer Berlin Heidelberg, 2010, pp. 65–74, http://dx.doi.org/10.1007/
978-3-642-12538-6 6.
[118] Y. Tan, Y. Zhu, Fireworks algorithm for optimization, in: Y. Tan, Y.
Shi, K. Tan (Eds.), Advances in Swarm Intelligence, vol. 6145 of Lect.
Note. Comput. Sci., Springer, 2010, pp. 355–364, http://dx.doi.org/10.1007/
978-3-642-13495-1 44.
[119] P. Fleming, R. Purshouse, Evolutionary algorithms in control systems engi-
neering: a survey, Cont. Eng. Pract. 10 (2002) 1223–1241.
[120] H.-S. Shih, W.-Y. Lin, E. Lee, Group decision making for topsis, in: The Joint
9th IFSA World Congress and 20th NAFIPS International Conference, 2001,
pp. 2712–2717, http://dx.doi.org/10.1109/NAFIPS.2001.943653.
[121] E. Burke, E. Hart, G. Kendall, J. Newall, P. Ross, S. Schulenburg, Hyper-
heuristics: an emerging direction in modern research technology, in: F.
Glover, G. Kochenberger (Eds.), Handbook of Metaheuristics, Kluwer, New
York, 2003, pp. 457–474.