Biological-based genetic algorithms for optimized disaster response resource allocation

https://doi.org/10.1016/j.cie.2014.05.001Get rights and content

Highlights

  • A biological-based generic algorithm for optimizing disaster resources allocation.

  • A case study of refuge site staff allocation and relief supply distribution plans.

  • A method of facilitating decision makers in optimizing resource use in disaster response.

Abstract

An effective disaster response requires rapid coordination of existing resources, which can be considered a resource optimization problem. Genetic algorithms (GAs) have been proven effective for solving optimization problems in various fields. However, GAs essentially use generation succession to search for optimal solutions. Therefore, their use of reproduction, crossover, and mutation operations may exclude optimal chromosomes during generation succession and prevent full use of previous search experience. Meanwhile, premature convergence caused by inadequate diversity of chromosome populations limits the search to a local optimum. Genetic algorithms also incur high computational costs. The biological-based GAs (BGAs) proposed in this study address these problems by including mechanisms for elite reserve areas, nonlinear fitness value conversion, and migration. This study performed experimental simulations to compare BGAs with immune algorithms (IAs) and GAs in terms of effectiveness for allocating disaster refuge site staff and for planning relief supply distribution. The simulation results show that, compared to other methods, BGAs can compute optimal solutions faster. Therefore, they provide a more useful reference when performing the decision-making needed to solve disaster response resource optimization problems.

Introduction

Since ancient times, natural disasters have threatened lives and property worldwide. Dilley et al. (2005) noted that, of the total population and land in Taiwan, 73% are potentially vulnerable to at least three different natural disasters (Dilley et al., 2005). This percentage is among the highest of all regions in the world. Thus, disaster management is a major concern of the government and citizens of Taiwan. Of the four stages of disaster management, including mitigation, preparedness, response, and recovery (McLoughlin, 1985), news media and education have achieved in a relatively high public awareness of the disaster mitigation and disaster preparedness stages.

However, the potential overload of information provided when a disaster affects numerous and widely distributed areas may be difficult to interpret. Minimizing loss of life and property requires an effective disaster response, including timely relief supply delivery, effective relief human resources deployment, and efficient delivery of human resources and supplies to affected populations (Tzeng, Cheng, & Huang, 2007). Planning and coordinating the use of existing relief resources and immediately assisting residents in disaster areas within the shortest possible time are major challenge for decision-makers. Therefore, effective relief resource planning is a critical disaster response issue.

The primary aims during the response phase are both rescue from immediate danger and stabilization of the condition of survivors. Tasks include relief, emergency shelter and settlement, emergency health care, water and sanitation, and tracing and restoring family links (Wex, Schryen, Feuerriegel, & Neumann, 2013). In the recovery phase, tasks are related to person finding, (ex-post) data analysis, intelligent infrastructure repair and the provision of various emergency services as well as resources in order to recover the critical infrastructure.

The quality of the relief efforts can be improved by effectively using the available technical resources. Because time, quantity and quality of the resources are the limited factors, emergency managers do have to find an optimal schedule for assigning resources in space and time to the affected areas. However, since it is difficult to assess and process all incoming information in an adequate manner, this problem is hard to solve (Fiedrich, Gehbauer, & Rickers, 2000). In particuar, two actions are needed after the disaster happens: “refuge site staff allocation planning” and “relief supply distribution planning”. The first action focuses on facilitating victims’ evacuation and on-site healthcare. The other action is for optimizing a balance among time constraints, emergency supplies and vehicles.

The efficient coordination of resources needed for an effective disaster response can be considered a resource optimization problem. Genetic algorithms (GAs) are evolutionary and feasible solutions to optimize complex problems (Holland, 1975). An effective disaster response requires decision-makers (DMs) to coordinate resource deployment efficiently and under time restraints. Therefore, a decision-making GA is proposed for supporting DMs in optimizing the coordination and deployment of disaster response resources under time constraints to minimize loss of life and property caused by a disaster.

The use of GA for efficiently comparing the advantages and disadvantages of chromosomes requires a function for calculating fitness values. Thus, the fitness function calculations of GAs incur a high computational cost (Cooper and Hinde, 2003, Povinelli, 2000, Povinelli and Feng, 1999). Another limitation of GAs is their relatively long execution times. Therefore, in addition to improving optimal solution search capability of the GA, the critical tasks are reducing fitness function calculation costs and improving time-efficiency.

A noted limitation of GAs is the use of roulette wheel selection so that chromosomes with high fitness values perform the preproduction needed for the next generation, including crossover and mutation operations or calculations for generating new chromosome populations and for replacing old chromosome populations. This limitation impairs performance because, by eliminating chromosomes with superior fitness values during generational crossover and mutation (Licheng & Lei, 2000), superior chromosome populations generated by previous search experiences may not be efficiently used for further searches.

Another limitation is that the high population diversity of chromosomes requires extensive searches for GAs in problem spaces. In the search for local optimal solutions over several generations of evolution, local optimal-solution-dominant chromosomes can limit chromosome population diversity, which confines GAs to local optimal solutions with no means of broadening the search to global optimal solutions. Over several generations of evolution, this phenomenon, which is known as premature convergence, can cause local optimal-solution-dominant chromosomes to limit chromosome population diversity and can confine GAs to local optimal solutions (Jung, 2009, Lozano et al., 2008, Taejin and Kwang Ryel, 2010, Tzeng et al., 2007).

Therefore, a biological-based genetic algorithm (BGA) is proposed for improving the solution performance of GAs by improving their optimal solution search capability and their execution time. Application of the proposed BGA can minimize loss of life and property by improving decision-making in disaster response. The BGA is characterized by the following features.

  • (1)

    After fitness values are adjusted, nonlinear fitness value conversion retains superior chromosomes by increasing their fitness values and eliminates inferior genetic groups by decreasing their fitness values. The fitness value conversion feature therefore increases competitive pressure within the genetic population.

  • (2)

    Elite reservation areas retain superior chromosomes for use in further crossover and mutation operations.

  • (3)

    By simultaneously storing the fitness values of retained superior chromosomes, elite reservation areas simplify the fitness function calculations.

  • (4)

    When generational evolution and reduced search capability cause premature convergence towards uniformity in chromosome populations, this method increases chromosome population diversity and moves beyond local optimal solutions by imitating the biological phenomenon of migration.

The remainder of this paper is structured as follows: Section 2 reviews the extant literature on disaster response resource planning and introduces the two major optimization algorithms compared in this study, i.e., GAs and immune algorithms (IAs); Section 3 details the proposed BGA; Section 4 discusses the experimental results; Section 5 concludes the study.

Section snippets

Current literature in disaster response resource planning

Recent studies in disaster response resource planning include Tzeng et al. (2007), who developed a disaster response model for planning three aspects of supply distribution: economics (minimizing total cost), effectiveness (minimizing total travel time), and equity (maximizing fairness) (Tzeng et al., 2007). Fuzzy multi-objective linear programming was also applied. Finally, a case study of the 9–21 Earthquake was used to perform drills and to recommend modifications needed when using the model

Biological-based genetic algorithm

Three mechanisms differentiate the proposed BGA from the conventional GA: an elite reserve area, non-linear fitness value conversion, and migration.

Experimental results

Experimental simulations were performed to compare the GAs, IAs, and the proposed BGA in terms of efficiency and effectiveness in planning the response to two problems that occur during a disaster: refuge site staff allocation and relief supply distribution.

Note that the CPLEX solver is not considered in this paper because the prior researches have shown that it has some weaknesses and limitations. For example, in contrast to GA, CPLEX needs more memory size and requires higher computing cost.

Conclusions

This study modified the conventional GA by including elite reserve areas, non-linear value conversion, and migration mechanisms. The resulting BGAs were then used to solve the problems of coordinating and distributing existing resources during a disaster response. The scenarios considered in the simulation experiments performed in this study included planning for the allocation of refuge site staff and for the distribution of relief resources. The experimental results show that the BGA is

References (40)

  • J.C. Bean

    Genetic algorithms and random keys for sequencing and optimization

    ORSA Journal on Computing

    (1994)
  • J. Cooper et al.

    Improving genetic algorithms’ efficiency using intelligent fitness functions

  • C. Darwin

    The origin of species by means of natural selection

    (1859)
  • L.N. de Castro et al.

    Learning and optimization using the clonal selection principle

    IEEE Transactions on Evolutionary Computation

    (2002)
  • M. Dilley et al.

    Natural disaster hotspots: A global risk analysis

    (2005)
  • J. Duda et al.

    A genetic algorithm for lot sizing optimization with a capacity loading criterion

    IEEE Congress on Evolutionary Computation

    (2007)
  • T.H. Emigh

    A comparison of tests for Hardy–Weinberg equilibrium

    Biometrics

    (1980)
  • H. Feltl et al.

    An improved hybrid genetic algorithm for the generalized assignment problem

  • L. Furong et al.

    Survey of artificial immune system

  • M. Gen et al.

    Genetic algorithms and engineering design

    (1997)
  • Cited by (21)

    • Sensor tasking for unmanned aerial vehicles in disaster management missions with limited communications bandwidth

      2020, Computers and Industrial Engineering
      Citation Excerpt :

      Finally, conclusions and opportunities for further research are outlined in Section 7. Disaster management is a well studied problem, focusing largely on disaster response and resource allocation (Chou, Tsai, Chen, & Sun, 2014; Rawls & Turnquist, 2010; Sarma, Bera, & Das, 2019). Our problem is largely a sensor tasking and scheduling problem; however, there are several components of our problem which extend traditional sensor scheduling problems.

    • A shortage risk mitigation model for multi-agency coordination in logistics planning

      2020, Computers and Industrial Engineering
      Citation Excerpt :

      Resource allocation models have proven useful to introduce human and material resources in the formulations with multiple participants. For instance, there are articles coordinating helicopter rescue tours minimizing cost (Barbarosoglu, Ozdamar, & Cevik, 2002), the distribution of heavy equipment minimizing total travel time (Chen, Peña-Mora, & Ouyang, 2011), the delivery of international resources minimizing cost (Adivar, Atan, Sevil Oflaç, & Örten, 2010), the allocation of disaster refuge staff minimizing total weighted distance (Chou, Tsai, Chen, & Sun, 2014), and allocating teams of volunteers to tasks minimizing cost (Falasca & Zobel, 2012). Rodríguez-Espíndola, Albores, and Brewster (2018b) focus on facility location, stock prepositioning, and relief distribution.

    • Implementing the NSGA-II genetic algorithm to select the optimal repair and maintenance method of jack-up drilling rigs in Iranian shipyards

      2020, Ocean Engineering
      Citation Excerpt :

      A wide range of studies has been performed in the field of drilling, repair and maintenance, and a variety of methods. Most of them have focused on how to choose the R&M method using AHP fuzzy analysis, neural networks and genetic algorithms in different fields (Jalali et al., 2020; Taghizade et al., 2019; Niu et al., 2018; Yazdi et al., 2019; Motamed and Majrouhi, 2018; Camara et al., 2018; Fu et al., 2018; Liao et al., 2018; Jacyna-Golda and Izdebski, 2017; Apergis et al., 2017; Gourlis et al., 2016; Pui et al., 2017; Wu et al., 2017;Bhandari et al., 2016; Abbassi et al., 2016; Chou et al., 2014; Xian-gang et al., 2013; Ahmadi et al., 2013; Bassi et al., 2012; Yu-guang et al., 2012; Hassan, 2012; Arunraj and Maiti, 2010; Alborzi, 2009 13–14; Shafieefar and Rezvani, 2007; Konak et al., 2006). In addition to research surveyed, the purpose of this study is to provide a method that can provide the most appropriate method of maintenance for masts.

    • Integrated optimal scheduling of repair crew and relief vehicle after disaster

      2019, Computers and Operations Research
      Citation Excerpt :

      Chou et al. (2014) developed Biological-based Genetic Algorithms (BGA) dealing with resource allocation problems in disaster response. The BGA developed by Chou et al. (2014) outperformed other solution procedures in cases of the allocation problem. Zhou et al. (2017) developed the emergency resource scheduling problem for a case of multiple periods in which unsatisfied demand and risk for choosing the destroyed road were minimized and optimal roads to rescue were efficiently selected.

    • Dynamic crowd evacuation approach for the emergency route planning problem: Application to case studies

      2018, Safety Science
      Citation Excerpt :

      In addition, the AIS algorithm has been proven to solve a wide range of complex and combinatorial problems such as optimization, data mining, computer security, and robotics (Freschi and Repetto, 2005; Aickelin and Greensmith, 2007; Bagheri et al., 2010; Dasgupta et al., 2011). Despite AIS adoption in other areas of emergency management (e.g. logistic management (Chou et al., 2014)), its adoption to the ERP problem is scarce. Consequentially, these features motivate the adoption of AIS algorithm as the proposed ERP approach.

    View all citing articles on Scopus

    This manuscript was processed by Area Editor Mitsuo Gen.

    View full text