DIEGO EVANDRO MAZZUCO
OTIMIZAÇÃO ADAPTATIVA BASEADA EM SIMULAÇÃO PARA PROBLEMAS DE ROTEIRIZAÇÃO E PROGRAMAÇÃO DE
VEÍCULOS COM JANELAS DE TEMPO
Dissertação submetida ao Programa de Pós-Graduação em Engenharia de Produção da Universidade Federal de Santa Catarina para a obtenção do Grau de Mestre em Engenharia de Produção.
Orientador: Prof. Dr. Enzo Morosini Frazzon
FLORIANÓPOLIS 2018
DIEGO EVANDRO MAZZUCO
OTIMIZAÇÃO ADAPTATIVA BASEADA EM SIMULAÇÃO PARA PROBLEMAS DE ROTEIRIZAÇÃO E PROGRAMAÇÃO DE
VEÍCULOS COM JANELAS DE TEMPO
Florianópolis, 28 de agosto de 2018.
________________________ Profa. Lucila Maria de Souza Campos, Dra.
Coordenador do Curso
Banca examinadora:
________________________ Prof. Enzo Morosini Frazzon, Dr.
Orientador
Universidade Federal de Santa Catarina (Videoconferência)
________________________ Prof. Antonio Cezar Bornia, Dr.
Presidente da Banca
Universidade Federal de Santa Catarina
________________________ Prof. Alexandre Hering Coelho, Dr. Universidade Federal de Santa Catarina
_______________________ Prof. Antônio Sérgio Coelho, Dr. Universidade Federal de Santa Catarina
Dedico este trabalho a todos da minha família, que sempre dão uma palavra de incentivo me apoiando em qualquer circunstância.
AGRADECIMENTOS
Primeiramente a Deus por iluminar sempre o meu caminho!
A minha família, por sempre acreditarem em mim, me ampararem em momentos difíceis e me darem apoio incondicional em todos os momentos da minha vida!
A todos os membros do ProLogIs por toda a troca de experiência vivida dentro e fora do laboratório. Fiz muitas amizades e com certeza levarei para a vida toda!
Aos professores do PPGEP e em especial ao meu orientador Prof. Dr. Enzo Morosini Frazzon, por toda a paciência e ajuda prestada para que eu pudesse chegar até aqui e também pela oportunidade de intercâmbio realizada durante o mestrado!
A todos os colegas e amizades que fiz durante meu intercâmbio na Alemanha no Research for Production and Logitics – BIBA Institute, agradeço todo o suporte desde o primeiro dia que cheguei em Bremen, além também das trocas de experiências e contribuições profissionais. A UFSC por todas as histórias vividas dentro de uma das melhores universidades do Brasil, durante esse mestrado!
Por fim, agradeço a todos que contribuíram de alguma forma para que eu conseguisse chegar até aqui!
“A vida começa todos os dias ” Érico Veríssimo
RESUMO
Organizações necessitam transportar seus produtos, para que cheguem no tempo certo, na medida certa e com a qualidade até o seu destino final. A qualidade define-se como a adequação do produto ao desejo do cliente, gerando satisfação, a qual, é almejada pelas empresas, já que assim poderão manter a competitividade e sua permanência no mercado. Um dos fatores mais importante com relação aos custos logísticos na cadeia de suprimentos é o custo de transporte. A programação de transporte, em cadeias de suprimento se caracteriza, tanto no âmbito empresarial quanto em pesquisas científicas, como um desafio relevante. Dessa forma esse trabalho tem como objetivo geral desenvolver uma abordagem de otimização baseada em simulação (simulation-based optimization – SBO) com enfoque em minimização de custo para problemas de roteirização de veículos com janelas de tempo. Neste trabalho estão presentes três artigos submetidos a periódicos ou conferências internacionais. O primeiro artigo apresenta uma análise do estado da arte em relação a otimização baseada em simulação para problemas de roteirização de veículos com o intuito de verificar os principais conceitos, direções e lacunas de pesquisas, e discute sobre o desenvolvimento de um novo conceito SBO em logística e transporte. O segundo artigo segue na direção de apresentar um modelo conceitual com a descrição do comportamento do sistema considerando a interação do método SBO com o nível de execução baseado numa sequência de atividades apresentandos em um framework, incluindo sua aplicação em um caso teste através de dados determinísticos. Por fim o terceiro artigo contempla a integração do modelo de simulação e otimização. Para resolver o modelo, um algoritmo baseado na meta-heurística Simulated Annealing (SA) foi desenvolvido e aplicado em um caso teste baseado em uma das instâncias fornecidas por Solomon, aplicado a 100 clientes atendidos por um depósito. Os veículos possuem um comportamento dinâmico, pois o tempo de viagem segue uma distribuição probabilística, que simula tempos estocásticos, assim como o tempo de atendimento em cada cliente. Por fim, os resultados são comparados com outros métodos apresentados na literatura. Do ponto de vista prático, o método proposto apresenta-se como suporte à tomada de decisão a gestores através de uma ferramenta que auxilia a programação de transportes. O presente estudo complementa pesquisas científicas contemporâneas e contribui para a inovação tecnológica direcionada ao aprimoramento da programação e controle de sistemas de transporte e logística, através do desenvolvimento de um método de otimização
baseado em simulação e orientado a dados para problemas de roteirização de veículos com janelas de tempo.
Palavras-chave: Problema de Roteirização de Veículos com Janelas de Tempo, Otimização, Simulação, Programação de Transportes, Logística.
RESUMO EXPANDIDO
OTIMIZAÇÃO ADAPTATIVA BASEADA EM SIMULAÇÃO PARA PROBLEMAS DE ROTEIRIZAÇÃO E PROGRAMAÇÃO DE
VEÍCULOS COM JANELAS DE TEMPO
INTRODUÇÃO
Organizações necessitam transportar seus produtos, para que cheguem no tempo certo, na medida certa e com a qualidade até o seu destino final. A qualidade define-se como a adequação do produto ao desejo do cliente, gerando satisfação, a qual, é almejada pelas empresas, já que assim poderão manter a competitividade e sua permanência no mercado. Um dos fatores mais importante com relação aos custos logísticos na cadeia de suprimentos é o custo de transporte. A programação de transporte, em cadeias de suprimento se caracteriza, tanto no âmbito empresarial quanto em pesquisas científicas, como um desafio relevante.
Um dos fatores mais importantes em termos de custos logísticos na cadeia de suprimentos é o custo de transporte. O transporte de carga representa o elemento mais importante em termos de custos logísticos para muitas organizações podendo absorver de um a dois terços dos custos logísticos totais. A programação de transporte, em cadeias de suprimento se caracteriza, tanto no âmbito empresarial quanto em pesquisas científicas, como um desafio relevante. No ambiente dinâmico dos sistemas produtivos a logística de cargas está sujeita a oscilações tanto de origem interna quanto externa. Dessa forma, uma nova abordagem que possibilite revisões em tempo real das programações de transporte à medida que as operações ocorrem, constitui uma oportunidade de pesquisa com potencial impacto prático. Sistemas de transporte são influenciados por uma ampla gama de influências estocásticas, tendo como consequência, que alguns parâmetros do problema de programação precisam ser expressos em termos de funções de probabilidades.
Na literatura recente, sugere-se que problemas estocásticos complexos, que ocorrem em sistemas de transporte, podem ser resolvidos via otimização baseada em simulação (SBO). Esta abordagem combina o poder de heurísticas de otimização com as vantagens de modelos de
simulação que, por sua vez, podem avaliar o efeito das mudanças de parâmetros até mesmo em sistemas muito complexos.
OBJETIVOS
O objetivo geral deste trabalho é propor um método adaptativo de otimização baseada em simulação para problemas de roteirização de veículos com janelas de tempo, adequado para lidar com sistemas envolvendo um grande número de participantes e interligações e cuja operação esteja sujeita a comportamentos estocásticos de origem interna e externa.
Os objetivos específicos deste trabalho são:
Realizar análise bibliográfica e bibliométrica das palavras-chave e temas envolvidos;
Propor um modelo conceitual do método adaptativo de otimização baseada em simulação para problemas de roteirização de veículos com janelas de tempo.
Implementar o método em um caso piloto, concentrando-se na construção de casos testes derivados de pesquisa de campo, buscando evidenciar a aplicabilidade e os resultados potenciais de sua aplicação.
METODOLOGIA
Primeiramente, uma revisão bibliométrica e sistemática da literatura foram executadas com o objetivo de identificar as principais tendências e oportunidades na área.
Posteriormente, para alcançar o objetivo proposto da dissertação, foi utilizada uma abordagem metodológica baseada em modelagem e simulação. Modelagem e simulação permitem uma melhor compreensão do ambiente em questão, identificando problemas, formulando estratégias e oportunidades, apoiando e sistematizando o processo decisório. O método proposto nesta dissertação foi composto por cinco etapas: (i) concepção do modelo conceitual de um sistema de transporte a ser analisado; (ii) adequação do modelo conceitual ao contexto do estudo; (iii) elaboração do modelo de otimização e simulação; (iv) coleta e análise de dados; (v) teste e análise comparativa de resultados computacionais.
CONCLUSÕES
O objetivo geral deste trabalho foi propor um método adaptativo de otimização baseada em simulação para problemas de roteirização de veículos com janelas de tempo, adequado para lidar com sistemas envolvendo um grande número de participantes e interligações e cuja operação esteja sujeita a comportamentos estocásticos de origem interna e externa.
Em busca do objetivo mencionado, a primeira etapa do estudo apresentou uma revisão literária de inovação tecnológica para melhoraria na programação e controle de sistemas de transporte e logística através do desenvolvimento de um método de otimização baseada em simulação para problemas de roteirização de veículos. Um estudo bibliométrico foi realizado e apresentou uma análise aprofundada sobre o que há de mais recente na literatura em relação às principais metodologias e métodos de simulação e otimização aplicados a sistemas de execução logística. Nessa fase, dezesseis artigos mais importantes foram selecionados para uma análise aprofundada para apontar as principais direções de pesquisa e principais lacunas.
Em seguida, o trabalho foi em direção a apresentar primeiramente um modelo conceitual que correspondia a uma descrição do comportamento do sistema considerando apenas uma fração das variáveis originais, derivadas em um modelo analítico apresentado passo a passo em um framework, aplicando o método de otimização baseado na meta-heurística têmpura simulada simulado em um caso de teste através de dados determinísticos. Posteriormente, a continuidade do estudo foi realizada considerando um ambiente estocástico a ser resolvido pela abordagem de otimização baseado em simulação, sendo aplicado em um caso teste com base em 100 clientes atendidos por um depósito. Os veículos tiveram um comportamento dinâmico, pois seu tempo de viagem segue uma distribuição probabilística, que simula tempos estocásticos, assim como o tempo de atendimento em cada cliente. Finalmente, os resultados foram comparados com outros métodos apresentados na literatura.
Como futuras oportunidades de pesquisa, uma pesquisa mais extensa e uma análise aprofundada sobre abordagens híbridas em logística e transporte podem ser feitas. Além disso, abordagens mais clássicas podem ser consideradas para investigar o impacto que outras práticas podem ter sobre o desempenho operacional da programação de transporte aplicada em problemas de roteamento de veículos.
PALAVRAS-CHAVE: Problema de Roteirização de Veículos com Janelas de Tempo, Otimização, Simulação, Programação de Transportes, Logística.
ABSTRACT
Organizations need to transport their products, to allow it to reach the client at the right time, in the right measure and with quality until the final destination. Quality is defined as the suitability of the product to the customer´s need, generating satisfaction, which is desire by the companies, since this way they can maintain the competitiveness and their permanence in the market. One of the most important factors regarding logistics costs in the supply chain is the transport cost. The transport scheduling in distributed production systems is a major challenge, in industrial praxis as well as in scientific research. In this way, this work has as general objective develop a simulation-based optimization (SBO) approach with focus on cost minimization for vehicle routing and scheduling problems with time windows. In this work three papers submitted to journals or international conferences are present. The first paper presents a bibliometric analysis and bibliographic review on simulation-based optimization for vehicle routing problems in order to verify the main concepts, directions and research gaps, and discusses the development of a new SBO concept in logistics and transportation. The second article follows in the direction of presenting a conceptual model with the description of the behavior of the system considering the interaction of the SBO method with the level of execution based on a sequence of activities presented in a framework, including its application in a test case through deterministic data. Finally, in the third article contemplates the integration of the simulation model and optimization. To solve the model, an algorithm based on the Simulated Annealing (SA) meta-heuristic was developed applied in a case test based on one of the instances provided by Solomon applied to 100 clients served by a depot. The vehicles have a dynamic behavior as their travel time follows a probabilistic distribution, which simulate stochastic times, as well as the service time in each customer. Finally, the results are compared with other methods presented in the literature.From a practical point of view, the proposed method presents itself as a support for the decision-making of managers through a tool that helps the scheduling of transport. The present study complements contemporary scientific research and contributes to technological innovation aimed to improving the scheduling and control of transport and logistics systems by developing a data-driven simulation-based optimization method for vehicle routing problems with time windows.
Keywords: Vehicle Routing Problem with Time Windows, Optimization, Simulation, Transport Scheduling, Logistic.
LISTA DE FIGURAS
Figure 1 – Research content ... 33
Figure 2 – Papers Procedure ... 34
Figure 3 - Methodology of the systematic literature ... 38
Figure 4 - Volume publication by year ... 39
Figure 5 – Most productive authors ... 40
Figure 6 – Journals with most publications ... 40
Figure 7 - Popular Keywords...41
Figure 8 - Interaction between adaptive heuristic optimization method and the Execution Level ... 63
Figure 9 - Approach Framework ... 64
Figure 10 - Simulated Annealing Algorithm ... 65
Figure 11 - Evolution graph ... 70
Figure 12 – Relative times to the route of each vehicle ... 71
Figure 13 - Simulated Annealing Framework ... 85
Figure 14 - Approach Framework ... 87
Figura 15 – Relative times to the route of each vehicle ... 89
LISTA DE TABELAS
Table 1 - Papers’ Framework ... 32 Table 2 - Papers submitted to an in-depth analysis ... 42 Table 3 - Main result of the experiment ... 71
LISTA DE ABREVIATURAS E SIGLAS
ACS Ant Colony System
BRAGECRIM Iniciativa Brasil-Alemanha para Pesquisa
Colaborativa em Tecnologia de Manufatura
CAPES Coordenação de Aperfeiçoamento de Pessoal de
Nível Superior
CPLS Cyber-physical Logistics System
CPS Cyber-physical System
DE Differential Evolution
GA Genetic Algorithm
HA Hybrid Approach
HDE Hybrid Differential Evolution
HMOEA Hybrid Multi-objective Evolutionary Algorithm
LES Logistics Execution System
SA Simulated Annealing
SBO Simulation-based Optimization
SCPS Socio-Cyber-Physical Systems
TSP Travelling Salesman Problem
TTRP Truck and Trailer Routing Problem
TTSMP The Tactical Time Slot Management Problem
VRP Vehicle Routing Problem
VRPBTW Vehicle routing Problem with Backhauls and Time Windows
VRPPD Vehicle Routing Problem with Pick up and Delivery VRPSPD Vehicle Routing Problem with Simultaneous Pick up
and Deliveries
SUMÁRIO 1. INTRODUCTION ... 27 1.1 OBJECTIVES...30 1.1.1 GENERAL OBJECTIVE ... 30 1.1.2 SPECIFIC OBJECTIVES ... 30 1.2 LIMITATIONS ... 30 1.3 DOCUMENT STRUCTURE ... 31 2. RESEARCH METHODS AND TECHNIQUES...31 2.1 EXPECTED RESULTS ... 34 REFERENCES ... 35
3. STATE OF THE ART IN SIMULATION-BASED
OPTIMIZATION FOR VEHICLE ROUTING PROBLEM...37 3.1 INTRODUCTION...37 3.2 METHODOLOGY...37 3.3 OVERVIEW OF THE PAPERS ... 38 3.4 SYSTEMATIC LITERATURE REVIEW ... 43 3.4.1 THE VEHICLE ROUTING PROBLEM - VRP ... 43 3.4.1.1 THE VRP WITH PICKUP AND DELIVERY AND TIME WINDOWS ... 43 3.5 OPPORTUNITIES ... 48 3.5.1 HYBRID APPROACH IN LOGISTIC AND TRANSPORT .. 48 3.5.2 SIMULATION-BASED OPTIMIZATION IN LOGISTICS .. 49 3.5.3 INDUSTRY 4.0 AND CYBER-PHYSICAL SYSTEMS ... 50 3.6 DISCUSSION OF THE PAPERS ... 51 3.7 CONCLUSION ... 55 3.8 REFERENCES ... 56 4. A CONCEPT FOR SIMULATION-BASED OPTIMIZATION IN VEHICLE ROUTING PROBLEMS ... 59 4.1 INTRODUCTION………..……..59 4.2 LITERATURE REVIEW ... 60 4.2.1 VEHICLE ROUTING PROBLEM WITH TIME WINDOWS 60
4.2.2 RESEARCH TENDENCIES IN LOGISTIC AND
4.3 CONCEPTUAL MODEL ... 61 4.4 APPROACH FRAMEWORK ... 63 4.4.1 OPTIMIZATION MODEL ... 65 4.4.1.1 PROBLEM DEFINITION ... 66 4.4.1.2 MATHEMATICAL MODEL ... 67 4.4.1.3 FORMULATION ... 68 4.5 TEST CASE ... 69 4.6 CONCLUSIONS ... 72 4.7 REFERENCES ... 72 5. AN ADAPTIVE SIMULATION-BASED OPTIMIZATION APPROACH APPLIED IN VEHICLE ROUTING PROBLEM WITH TIME WINDOWS ... 75 5.1 INTRODUCTION ... 75 5.2 LITERATURE REVIEW ... 77
5.2.1 OPTIMIZED PROGRAMMING IN DISTRIBUTED
PRODUCTION SYSTEMS ... 77 5.2.2 SIMULATION OF SYSTEMS... 78 5.2.3 OPTIMIZATION AND SIMULATION OF COMPLEX AND STOCHASTIC SYSTEMS – APPLICATIONS IN VRPTW ... 80 5.3 PROBLEM FORMULATION ... 82 5.3.1 SIMULATED ANNEALING META-HEURISTIC FOR THE VRPTW ... 84 5.3.2 APPROACH FRAMEWORK ... 86 5.4 TEST CASE - EVALUATION AND RESULTS ... 88 5.5 CONCLUSION ... 91 5.6 REFERENCES ... 91 DISCUSSION ... 97 CONCLUSIONS ... 101
27
1. INTRODUCTION
Organizations need to transport their products, so that the clients served at the right time, in the right measure and with quality. The satisfaction of the client, whether internal or external, is desired by the companies, since this way they can maintain the competitiveness and their permanence in the market. One of the most important factors regarding logistics costs in the supply chain is the transport cost. According to Ballou (2006), cargo handling represents the most important element in terms of logistics costs for many companies and can absorb one to two-thirds of the total logistics costs. Companies aware of this, and in search of sustainable and economic alternatives, are willing to perfect their means of taking their products to their customers.
The transport scheduling in distributed production systems is a major challenge, in industrial praxis as well as in scientific research. In the dynamic scenario of production systems, transport is subject to oscillations from internal and external factors increasing the complexity to solve the problems. In the recent literature, it is suggested that complex stochastic problems, which occur in transport systems, can be solved via simulation-based optimization (SBO) (Lin and Chen, 2015). This approach combines the capabilities of heuristics optimization with the advantages of simulation models that can evaluate the effect of parameter changes even in very complex systems. However, in addition to stochastic influences that can be described as functions of probability, real systems have a feature that is neglected, for the sake of simplicity, in most models: they are highly dynamic systems that have to adapt permanently to a variety of external influences such as unpredictable demand, high priority urgent requests, or disturbances such as vehicle breakdown.
The transport scheduling is directly related to determining the paths that a vehicle or fleet of vehicles must travel to optimize the process of collecting and delivering goods and/or people with the focus of minimizing the costs. This process involves restrictions, such as pick-up and delivery time window, demand, type of vehicle, employees, garage, availability of vehicles, among others (Li & Fu, 2002).
According to Bodin & Golden (1981) routing is the sequencing of points of pickup and/or delivery that a vehicle must go through
28
systematically, starting and finishing in a depot or customer related to the distribution of products and services. However, the transport programming is the insertion of conditioning factors of stop and arrival times in this sequencing, that is, are restrictions that are inserted that can involve the time of service in each client, travel times, the maximum journey of work, among others.
The first most famous routing problem is probably the Travelling Salesman Problem (TSP), which describes the task of finding the shortest tour through a given set of cities where the tour comprises each city exactly once (Ehm & Freitag, 2016). If it is in addition necessary to split the number of jobs into tours for several vehicles, it is called Vehicle Routing Problem (VRP). The TSP is a well-known optimization problem in operations research that has nowadays received much attention because of its practical applications in industrial and service problems. Different algorithms including exact, heuristics and metaheuristics have been explored during the decades (Cordeau et al., 2010; Yadlapalli et al, 2009). Although optimal solutions can be obtained using exact methods for the small size of TSP problems, the computational time required to solve adequately large problem instances is still impracticable, outstanding to the high computational cost and poor performance of exact methods in solving large problem instances. So, in line to data constraints and complexity, the most commonly used method for solving routing problems is based on heuristic and meta-heuristic algorithms, which perform a combinatorial analysis of the data to demonstrate an approximation of the optimal result (Wang et al, 2015). Several algorithms have been proposed in the literature to solve these problems in a near-optimal solution with reasonable computational times (Renaud et al., 1998).
The characterization of the problem to be studied in this dissertation is based on the vehicle routing problem. The vehicle routing problem is considered one of the most important problems in the wide range of operational research studies. To plan the routes of vehicles departing from a depot to conciliate with the specifications of customers distributed in a geographic area, focusing on the reduction of cost, time or distance, have been studied over the years, performing an important function in several domains, such as systems business, economic and
29
service sectors. To be more specific this study will be applied in a scenario of vehicle routing problem with the addition of a time window variant for each client, defined as vehicle routing problem with time windows (VRPTW). For this approach, the delivery of the products must occur within a period (ai, bi), in which ai and bi are the earliest and the latest
allowable times that the service should be taken place (Tavakkoli-moghaddam et al., 2011).
To solve the model, an algorithm based on the Simulated Annealing (SA) heuristic was developed. For Coppin (2004) meta-heuristics approaches are often considered part of the evolutionary or artificial intelligence class of the optimization methods. In this work, metaheuristics are going to be considered part of the optimization branch of the analytical methods, because besides the characterization by Coppin (2004) it is easy to notice that meta-heuristics are being used for searching for good solutions in complex cases of mathematical programming as logistic execution systems. Küçükoğlu et al. (2014) define SA as a stochastic method for solving combinatorial problems inspired by the metallurgy annealing process. SA works by emulating the physical process, in which a solid is heated to a high temperature and slowly cooled, allowing the solid to crystallize. The data collection used for the test case in this work, including the position of customers and depot, demands, time windows, and vehicle capacity, were taken from a benchmarking problem created by Solomon in 1987.
The complexity of real systems is related to a large number of agents and their interactions. Simulation-based techniques can be used both to develop and to evaluate the effect of parameter changes even on very complex systems, nevertheless, it cannot provide an efficient optimization. The current state of a system can be characterized by its structure, parameters, process probability distributions, among others. Thus, the simulation should be integrated with optimization in order to continuously improve the outputs under various environments and finally to optimize the system’s performance (Lin and Chen, 2015; Hu et a.l, 2008).
Therefore, the main question addressed in this dissertation is how a simulation-based optimization model could assist in vehicle routing and scheduling problems with time windows. An approach that enables
real-30
time review of transport schedules as the operations occur is a scientific challenge with potential practical impact.
1.1 OBJECTIVES
1.1.1 GENERAL OBJECTIVE
The general objective of this work is to propose an adaptive simulation-based optimization method for vehicle routing and scheduling problems with time windows, suitable for dealing with systems involving large numbers of participants and interconnections between participants and whose operation is subject to stochastic, internal and external origin.
1.1.2 SPECIFIC OBJECTIVES
The specific objectives of this work are:
To perform bibliographical and bibliometric analysis on the key words and themes involved;
To propose a conceptual model of the adaptive method of simulation-based optimization for vehicle routing problem with time windows.
To implement the method in a pilot case, concentrating on the construction of test cases derived from field research, seeking to evidence the applicability and potential results of its application.
1.2 LIMITATIONS
Some limitations shall be considered in the application of the present work.
The bibliographical and bibliometric analysis will consider only the documents written in English language, as well as only two databases: Web of Science and Science Direct.
No type of statistic will be presented in this work about which are the most used solving methods in the literature. The characterization of which ones are the most used ones will be done relying on other studies presented in the literature, for example, surveys of data collection and then one will be chosen that fits in a better way for the proposed approach.
31
The proposition of a mathematical model will be done looking for a suitable one that is already validated. The development of a new mathematical model is not in the scope of the present research.
1.3 DOCUMENT STRUCTURE
The present work is a compendium composed by three scientific papers submitted/approved to conferences and journals elaborated along the last two years. More information about the papers is shown in table 1. Session 3 presents the first paper from the table 1, and presents a bibliometric analysis and bibliographic review on simulation-based optimization for vehicle routing problems in order to verify the main concepts, directions, and research gaps, and discusses the development of a new SBO concept in logistics and transportation.
Session 4 brings the second paper on the table, and goes in the direction of presenting a conceptual model with the description of the behavior of the system considering the interaction of the SBO method with the level of execution based on a sequence of activities presented in a framework, having as solution the application in a test case through deterministic data.
Finally, session 5 brings the third paper, we gave continuity to the study integrating the model of simulation and optimization based on stochastic data being subject to disturbances which are solved in real time by the model. From a practical point of view, the proposed method presents itself as a support for the decision-making of managers through a tool that helps the scheduling of transport.
2. RESEARCH METHODS AND TECHNIQUES
In order to fulfill all the pointed objectives of the research previously described, a research was designed and the objectives and methods from each phase were divided into three main steps as shown in
32
Figure 1, each step executed was reported in one of the papers cited in Table 1.
Table 1 – Papers’ Framework
TITLE AUTHORS PUBLICATION
P
AP
E
R
1 State of the Art in Simulation-based Optimization for Vehicle
Routing Problem
Mazzuco, D.E.; Oliveira, D.L.; Frazzon, E.M. Published on International Journal of Production Management and Engineering P AP E R 2
A concept for simulation-based optimization in Vehicle Routing Problem
Mazzuco, D. M; Carreirão Danielli, A. M; Oliveira, D. L; Santos, P. P. P; Coelho, L. C; Frazzon, E. M Published on proceedings from 16th IFAC Symposium on Information Control Problems in Manufacturing P AP E R 3 An adaptive simulation-based optimization approach applied in Vehicle Routing Problems with Time
Windows
Mazzuco, D.E.; Frazzon, E.M.
Submitted to International Journal of
Production Research
Source: Author
Initially, in the first phase, the research had an exploratory character, since in the paper 1, the gaps that guided the research were investigated by the identification of the main simulation and optimization practices applied to vehicle routing problems. The research will be done in reference databases that is composed of a large number of quality journals on the subject. Data analysis software will be used with the intention of extracting statistics to verify tendencies and to make an optimized reading of the mass of extracted data so that better conclusions can be made and assist in the continuity of research. According to Andrade (2001), the exploratory research aims to provide more information about the subject, facilitating the delimitation of the topic to be worked on and directing the research that is to be developed.
33
Fig. 1 - Research content
Source: Author
In the second phase of the research, the paper 2 focused firstly to present a conceptual model that corresponds to a description from the “real system” which only an original variables fraction that defines the system behavior are considered, for then, derived from the conceptual model, to abstract an analytical mathematical model (relations of the system expressed by mathematical functions) and an experimental model of simulation (seeking to emulate by means of logical relations the functioning of the system). Based on the conceptual model, a framework with the techniques used was presented as well and then applied in a first test case with deterministic data.
Finally, paper 3, assumed quantitative research characteristics, working with numerical data which were analyzed by the analytic method presented to evaluate the system behavior. Quantitative research focuses on objectivity, that is, it seeks to translate all that can be quantifiable by measuring predetermined variables, which it is desired to verify and explain its influence on other variables (APPOLINÁRIO, 2006). Basically, quantitative research uses measurable data to formulate facts and uncover patterns in research.
34
Looking at it more broadly, the first step in paper one of the compendium aims to provide a better understanding of the research area, as its main concepts, main problems to be solved, as well as the main approaches to solving it. Step two in the second paper goes in the direction of a better understanding the environment in question, identifying problems, formulating strategies and opportunities, and presenting a conceptual model which corresponds to a description of the system behavior to abstract an analytical model shown in a framework of the approach. Step aims to support the application and validation of an adaptive approach based on a simulation-based optimization method to deal with the dynamic behavior of VRPTW comparing the results with others in the literature. The procedure is shown in Fig. 2:
Fig. 2 – Papers Procedure
Source: Author
2.1 EXPECTED RESULTS
At the end of the research, it is expected to obtain an iterative method that should be robust to the realization of an adaptive simulation-based optimization applied in a vehicle routing problem with time windows, through a data set of a transportation and logistics system.
35
3. REFERENCES
Andrade, M. M. de. Introdução à metodologia do trabalho científico: elaboração de trabalhos na graduação. 5. ed. São Paulo: Atlas, 2001.
Appolinário, F. Metodologia da ciência: filosofia e prática da pesquisa. São Paulo: Pioneira Thomson Learning, 2006.
Ballou, R. H. Gerenciamento de Cadeia de Suprimentos/ Logística Empresarial. Porto Alegre: Bookman, 2006.
Bodin, L.D., Golden, B. (1981) Classification in vehicle routing and scheduling. Networks, v.11, p.97-108.
Coppin, B. Artificial intelligence illuminated. Jones & Bartlett Learning, 2004. Cordeau, J. F., Dell’Amico, M., Iori, M. (2010). Branch-and-cut for the pickup and
delivery traveling salesman problem with FIFO loading, Computers & Operations Research, 37(5), 970-980. http://dx.doi.org/10.1016/j.cor.2009.08.003.
Ehm, J., Freitag, M. (2016). The benefit of integrating production and transport scheduling. 48th CIRP Conference on MANUFACTURING SYSTEMS - CIRP CMS 2015. Procedia CIRP 41, 585 – 590.
Hu, X., Li, Y., Guo, J., Sun, L., Zeng, A. Z. (2008) A Simulation Optimization Algorithm with Heuristic Transformation and its Aplication to Vehicle Routing Problems. International Journal of Innovative Computing, Information and Control. V. 4, Nº 5.
Küçükoğlu, I., Öztürk, N. (2013). A differential evolution approach for the vehicle routing problem with backhauls and time windows. Journal of Advanced Transportation, 10.1002/atr.1237.
Li, L.Y.O., Fu, Z. (2002) The school bus routing problem: a case study. Journal of the Operational Research Society, vol. 53, pp. 552–558.
Lin, J. T., & Chen, C. M. (2015). Simulation optimization approach for hybrid flow shop scheduling problem in semiconductor back-end manufacturing. Simulation Modelling Practice and Theory, 51, 100-114.
Renaud, J., Fayez, F., Boctor, F. (1998). An efficient composite heuristic for the symmetric generalized traveling salesman problem, European Journal of Operational Research, 108(3), 571-584. http://dx.doi.org/10.1016/S0377-2217(97)00142-2.
Tavakkoli-Moghaddam, R., Gazanfari, M., Alinaghian, M., Salamatbakhsh, A., & Norouzi, N. (2011). A new mathematical model for a competitive vehicle routing problem with time windows solved by simulated annealing. Journal of Manufacturing Systems, 30(2), 83–92
Wang, C., Mu, D., Zhao, F., Sutherland, J. W. (2015). A parallel simulated annealing method for the vehicle routing problem with simultaneous pickup-delivery and time windows. Computers & Industrial Engineering, 83, 111-122.
Yadlapalli, S., Malik, W.A., Darbha, S., Pachter, M. (2009). A Lagrangian-based algorithm for a Multiple Depot, Multiple Traveling Salesmen Problem, Nonlinear
36
Analysis: Real World Applications, 10(4), 1990-1999. http://dx.doi.org/10.1016/j.nonrwa.2008.03.014.
37
3. STATE OF THE ART IN SIMULATION-BASED
OPTIMIZATION FOR VEHICLE ROUTING PROBLEM 3.1 INTRODUCTION
Recent literature suggests that simulation-based optimization (SBO) is a powerful tool for solving complex stochastic problems (Lin and Chen, 2015). This approach combines the power of optimization heuristics with the advantages of simulation models, which in turn, can evaluate the effect of parameter changes even on very complex systems. However, as the simulation model is a direct description of a problem, this method provides only feasible, but not necessarily an optimal, solution under certain conditions. Thus, the simulation should be integrated with optimization in order to continuously improve the outputs under various environments and finally to optimize the system’s performance (Hu et al, 2008).
Finding an optimal route for vehicles delivering products to customers is one of the key functions in any logistics systems. Since the first works proposed in the literature in late 1959, the literature on vehicle routing problems have been highly increasing. A lot of works, as well as state-of-the-art surveys, have been published (Montoya-Torres et al., 2015).
In the dynamic scenario of production systems, loading logistics is subject to both internal and external oscillations. In this way, a new approach that enables real-time revisions of transport schedules as operations occur constitutes a research opportunity with potential practical impact. The present study aims to research publications that apply to this new approach, aiming as result a theoretical foundation as to the creation to the simulation-based optimization in a case test as to future researchers.
3.2 METHODOLOGY
The objective of this chapter is to present the methodology used to search articles on the topic of SBO applied to logistic execution systems.
All the literature review was carried out based on "Web of Science" and “Science Direct” databases. The keywords identified as most relevant related to the scope of the article were "simulation-based optimization", "vehicle routing problem" and "hybrid approach". The keywords were combined with other ones such as “pick up and delivery”,
38
“time windows”, “meta-heuristic”, “transport scheduling”, "logistic" and "optimization". The fig. 3 bellow presents the methodology for choosing the articles.
Fig. 3 – Methodology of the systematic literature
Source: Author
The first selection was made based on the titles, being selected those related to delivery, simulations and optimization problems. The second selection was through the abstracts. The third selection was reading each paper introduction and conclusion analysis. Finally, the last step of the selection was the papers complete reading, resulting in sixteen papers utilized.
3.3 OVERVIEW OF THE PAPERS
The bibliometric review was studied using the total of papers selected after the secondary keywords. Figure 4 below presents the total publication of papers regarding the studied area over the years, from 1999 to 2017.
39
Fig. 4 – Volume publication by year
Source: Author
It suggests that the interest in routing problems has increased in the last years, having the best number of publications in 2016. Figure 5 presents the authors with most publications related to the studied area.
40
Fig. 5 – Most productive authors
Source: Author
Figure 6 presents journals and conferences that have more publications and interest in the studied areas.
Fig. 6 – Journals with most publications
41
From the report, it is possible to obtain which keywords are most similar to those used in the search showed in chapter 2. Figure 7 shows these keywords.
Fig. 7 – Popular Keywords
Source: Author
The literature review starts with the combination of keywords that is pertinent in the study to find papers at the databases. But with the combination of the keywords, VRP, VRPPD, VRPTW, cyber-physical systems (CPS), SBO and Hybrid Approach (HA) it was not possible to find many results related with the themes, hence, the search has combined
42
these words in pairs to search the results more often. Below there are the 16 most important papers found in the literature regarding the keywords of the bibliometric analysis. Table 2 presents these papers.
Table 2: Papers submitted to an in-depth analysis
Year Title Author
2017 Hybrid Differential Evolution Optimization for the Vehicle Routing Problem with Time Windows and Driver-Specific Times
(Pu et al)
2016 A hybrid heuristic approach for the multi-commodity pickup-and-delivery traveling salesman problem
(Hernández-Pérez et al)
2016 Vehicle Routing Problem with Time Windows based on
Two-stage Optimization Algorithm (Cai et All) 2015 Simulation optimization approach for hybrid flow shop
scheduling problem in semiconductor back-end manufacturing
(Lin & Chen)
2013 A hybrid meta-heuristic for multi-objective vehicle
routing problems with time windows (Baños et al) 2013 Towards socio-cyber-physical systems in production
networks
(Frazzon et al) 2013 Cyber-physical logistics system-based vehicle routing
optimization
(Lai et al) 2012 A genetic Algorithm based approach to VRP with
simultaneous pick-up and deliveries
(Tasan & Gen)
2012 A Cyber-Physical Future (Rajkumar)
2012 Logistics Distribution Vehicle Routing Problem with Time Windows
(Ailing,) 2007 An Improved Evolutionary Algorithm for Dynamic
Vehicle Routing Problem with Time Windows (Wang et al) 2007 General solutions to the single vehicle routing problem
with pick-ups and deliveries
(Gribkovskai et al) 2006 A hybrid multiobjective evolutionary algorithm for
solving VRP with time windows
(Tan, Chew, & Lee) 2000 Optimal Production-distribution planning in supply chain
management using a hybrid simulation-analytic approach
43
2002 Parallel simulated annealing for the vehicle routing
problem with time windows (Czech Czarnas) and 1996 Heuristic approaches to vehicle routing with backhauls
and time windows
(Thangiah et al)
Source: Author
3.4 SYSTEMATIC LITERATURE REVIEW
In this section, a relevant theory will be presented to elucidate the state of the art regarding Vehicle Routing Problem and the variants Vehicle Routing Problem with Pick up (VRPPD) and Delivery and also Vehicle Routing Problem with Time Windows (VRPTW).
3.4.1 THE VEHICLE ROUTING PROBLEM - VRP
Since Dantzig and Ramser (1959) introduced the VRP, it has been one of the most widely analyzed NP-hard problems. The main purpose of the vehicle routing problem (VRP) is to deliver a set of customers with known demands on minimum-travel routes and terminating at the same depot. The VRP is defined on an undirected network G = (V,E) with a vertex set V = {1, 2,…, n} and an edge set E. (Pérez-Rodríguez et al., 2015).
Each other vertex i>1 represents a customer with a known service time si, and each edge has a non-negative travel time tij= tji. The VRP
consists of determining a set of m vehicle trips of minimal total time, such that each vehicle starts and ends at the depot, each customer is visited exactly once, and the total demand handled by any vehicle does not exceed the vehicle’s capacity Q (Pérez-Rodríguez et al., 2015).
The basic VRP has been extended to include aspects such as the characteristics of the network, the fleet, and the customers, making the problem more difficult to be solved (Bochtis & Sørensen, 2009 ).
3.4.1.1 THE VRP WITH PICKUP AND DELIVERY AND TIME WINDOWS
A classification of VRP with pick-up and delivery has been presented in the literature review paper of Parragh, Doerner, and Hartl (2008). They first define a main class of problems, which deal with the transportation of goods from the depot to the customers and from the
44
customers to the depot and then the second main class is identified as classical to all those types of problems where goods are transported between pick-up and delivery points may be paired or unpaired.
The single-vehicle routing problem with pickups and deliveries (SVRPPD) is defined as follows. Let G = (V,A) be a graph where V = {0,1, ... ,n} is the vertex set, A ={(i, j): i, j V, i j} is the arc set, and C = is a cost matrix defined on G. Vertex 0 is a depot at which is based a vehicle of capacity Q, while the remaining vertices represent customers. With each vertex i V\{0} are associated a non-negative pickup demand 𝑝𝑖 and a non-negative delivery demand 𝑑𝑖, with
𝑝𝑖 + 𝑑𝑖 > 0. It is assumed that and for otherwise
the problem is infeasible. In practice, the products to be picked up are different from those delivered (Gribkovskaia et al., 2007).
The formulation and notation bellow was proposed by Tasan and Gen (2012) and adapted, using the formulation proposed by Luo et al. (2015), based on the paper scenario and objectives.
Notations:
V: Set of vehicles; J: Set of customers;
𝑱𝟎: Set of Customers including depot; A: Set of arcs;
𝑲𝒗: Vehicle capacity; n: Number of nodes;
𝒄𝒊𝒋: Distance between nodes i and j, i≠j; 𝑝𝑗: Pick-up amount of customer node j;
𝒅𝒋: Delivery amount demanded by customer
node j;
𝑥𝑖𝑗𝑣: Binary decision variable that
indicates whether 𝑣𝑡ℎ vehicle travels from
i to j; (𝒊, 𝒋)𝒗: Arc to be traveled by vehicle v, from
node i to node j;
𝑙′𝑣: Load of 𝑣𝑡ℎ vehicle when leaving the
depot; 𝒍𝒋: Load of a vehicle after served customer
node j;
𝑠𝑗 Variable used to avoid subtours, can be
interpreted as position of node j J in the route;
𝒘𝒗= 𝟏 if vehicle v is used and 𝑤𝑣= 0 if it
45
M: Sufficiently large number [e.g. M can be calculated as the maximum value of either the total customer delivery and pick-up demands of the distances between each nodes as given in (1)].
(1) Then, the formulation is as follows:
Minimize (2) Subject to (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)
46
The main objective (2) is to minimize the total set of used vehicles, the total distance traveled and, consequently, the travel total cost. Constraint (3) ensures that each customer node is served exactly once. Constraint (4) guarantees that for each customer node, the same vehicle arrives at and leaves this node. Initial vehicle loads are calculated as in (5), each vehicle’s initial load is the accumulated demand of all customer nodes assigned to this vehicle. (6) Balances a load of vehicles after vehicles visit the first customer node on their route. For other customer nodes, loads of vehicles are calculated as in (7) similarly through their routes. (8) And (9) impose capacity constraint (11) maintains non-negativity of 𝑠𝑗 and 𝑥𝑖𝑗𝑣 is a binary decision variable as in
(12). Finally, a non-negative requirement (13) to the number of vehicles used.
Regarding the Vehicle Routing Problem with Time Windows, Ailing (2012) presents in his paper the mathematical model that the objective is to determine a vehicle route plan which minimizes the total cost with the following constraints:
Each route starts and ends at the central depot;
The quantity of goods delivered must not exceed the capacity of the vehicle;
The total length of each route must not exceed the constraint; Each vehicle must visit the customer within the time fixed by the
customer;
Each customer is served exactly once by exactly one vehicle. Assume that the central depot is node 0, and N customers are to be served by K vehicles. The demand of customer i is iqi 1,2,…, N; the capacity of vehicle k is , Qk k = 1,2,…, K and the maximum allowed travel distance by vehicle k is , Dk k= 1,2,…, K; 𝑋𝑖𝑗𝑘: Binary variable that
indicates if vehicle k travels the route between clients i and j; 𝑌𝑗𝑘: Binary variable that indicates if vehicle k visits the client j. Cijk is the cost of
traveling from customer i to customer j by the vehicle k . d kij is the travel distance from customer i to customer j by the vehicle k ; Tij（i, j ϵ{0,1,…, N}, i ≠ j）is the time from customer i to customer j; ski is the time vehicle k visit customer i, if vehicle k don’t visit customer i, ski indicates none; [ai, bi] is the fixed time window serving for customer i, which specifies that vehicle k must visit customer i within the time from ai to bi, if vehicle k reaches customer i earlier than ai, then the vehicle must wait; if vehicle
47
k reaches customer i later than bi, vehicle k must be punished (AILING, 2012). Minimize (1) Acoording to: (2) (3) (4) (5) (6) (7) (8) (9) The objective function (1) is to minimize the total cost by all vehicles; Constraints (2), (3), (4) ensure that each customer is served exactly once; Constraint (5) ensures the limit of the total length of each route; Constraint (6) shows that the number of goods delivered must not exceed the capacity of the vehicle; Constraint (7) specifies the time window; Constraint (8) shows that vehicle k, can’t reach customer j before the time sk
i + Tij, where P is a large number; Constraint (9) ensures that
48
3.5 OPPORTUNITIES
3.5.1 HYBRID APPROACH IN LOGISTIC AND TRANSPORT
Although heuristic and metaheuristic algorithms can obtain better quality solution compared with the exact algorithm, many researchers have found that the employment of hybridization in optimization algorithms can improve the quality of the problem in comparison with these algorithms (Yousefikhoshbakht et al., 2016). Therefore, a description of hybrid simulation/analytic models and the modeling is presented in this topic based on approaches developed.
A hybrid modeling consists of building independent analytic and simulation models of the total system, developing their solution procedures, and using their solution procedures together for problem-solving (SARGENT, 1994).
Haridass et al. (2013) developed a simulated annealing algorithm that interacts with a deterministic simulation model. The simulated annealing algorithm consists of two stages. In the first stage, an initial solution is made feasible by meeting all demands for the day. The solution is then improved using transfer, exchange, interchange and stop operators. Improvement is continued until the user-specified stopping temperature of the simulated annealing process is reached.
Lee and Kim (2000) propose a hybrid approach combining the analytic and simulation model. Operation time in the analytic model is considered as a dynamic factor and adjusted by the results from the independently developed simulation model, which includes general production-distribution characteristics. The study obtains the more realistically optimal production-distribution plans for the integrated supply chain system reflecting stochastic natures by performing the iterative hybrid analytic-simulation procedure.
The procedure of the hybrid simulation-analytic approach is based on imposing adjusted capacities derived from the simulation model results. The procedure consists of the following steps (H. Lee and Kim, 2000):
Step 1. Obtain production and distribution rates from the analytic model.
Step 2. Input production and distribution rates to the independently developed simulation model.
49
Step 4. If the results show that production and distribution can be produced and distributed within the capacity, then go to step 6. Otherwise, go to step 5.
Step 5. Adjust capacity constraints for the analytic model based on the simulation results from step 3 and regenerate production and distribution rates and go to step 6.
Step 6. Production and distribution rates given by the analytic model may be considered to be optimal solutions.
Step 7. Stop.
Yousefikhoshbakht et al. (2016) developed an approach to solve the Traveling Salesman Problem based on a hybrid meta-heuristic algorithm called REACSGA - reactive bone route algorithm that uses the ant colony system (ACS) for generating initial diversified solutions and the genetic algorithm (GA) as an improved procedure are applied. In this problem, a salesman starts to move from an arbitrary place called depot and after visits all of the nodes, finally comes back to the depot and the objective is to minimize the total distance traveled by the salesman. The computational result shows that the results of the proposed algorithm are competitive with other metaheuristic algorithms for solving the TSP in terms of better quality of solution and computational time respectively.
3.5.2 SIMULATION-BASED OPTIMIZATION IN LOGISTICS
A model is a representation of a situation or reality, as seen by a group of people, and built to assist the handling of the situation in a systematic way. Enables a better understanding of the context, to formulate strategies and opportunities to support and systematize the process of decisions (Atzori, Iera, and Morabito 2010). Historically, problem-solving in complex systems involves modeling techniques using heuristics, also, meta-heuristics and simulations (Longo, 2010).
Heuristic, are sets of steps, taken in sequence, to resolve a combinatorial optimization problem. Basic heuristics used in common logistics problems such as the vehicle routing problem can be described as either construction heuristics or local improvement heuristics (Griffis et al., 2012; Bräysy and Gendreau, 2005). Construction heuristics build a feasible solution using iterative steps based on some criteria such as minimizing cost, time or distance in a VRP case, for example. In contrast, local improvement heuristics stars with an initial solution and iteratively improve on it by considering neighboring solutions (Griffis et al., 2012). Meta-heuristics are defined as solution methods that orchestrate an
50
interaction between basic local improvement heuristics and higher level strategies aimed at escaping from local optimums in a solution space (Griffis et al., 2012; Glover and Kochenberger, 2003). Meta-heuristics differ from the classic heuristics because perform a much more thorough search of the solution space, allowing inferior and sometimes infeasible moves, as well as recombinations of solutions to create new ones (Griffis et al., 2012; Cordeau et al., 2002).
Therefore, should be analyzed which best simulator to be used to develop a robust optimization model capable of dealing with changes in system state, simulated in various scenarios. Finally, the performance of the adaptive method is evaluated through the test scenarios based on empirical data based on a real case, and after the evaluation results will be used for fine-tuning the optimization heuristics. As mentioned, simulation is a powerful tool for the analysis and evaluation of complex and stochastic systems such as contemporary manufacturing and transportation environments (J. T. Lin and Chen, 2015).
The complexity of most of the real-world systems is related to their stochastic nature as well as to a multitude of internal and external interactions. However, it cannot provide an efficient optimization of these systems in relation to one or more performance indicators (e. g. waiting times, production costs, etc.). Optimization methods are mainly used if a complex system can be modeled by a simplified abstraction. Therefore, both approaches individually are limited in optimal decision making for complex and stochastic systems, such as scheduling transport and logistics systems. A promising approach to combine the strengths of both is simulation-based optimization (SBO). In this scenario, the simulation model is used as the objective function of the optimization and the optimization method determines the simulation parameters (Krug Et Al., 2002). Aspects like the physical configuration or operational rules of a system can be considered. Its applications have grown in all areas, assisting managers in the decision making process and enabling a better understanding of processes in complex systems (O’Kane, Spenceley, and Taylor 2000). Simulation can already be used to study systems in the design stage (Banks et al., 2000).
3.5.3 INDUSTRY 4.0 AND CYBER-PHYSICAL SYSTEMS
With the emergence of the 4th Industrial Revolution and the consequent advance of automation, it is necessary for companies to
51
reconcile human work with new technologies, facilitating the work that previously required a lot of time from employees and demanded a high level of care due to the great difficulty. Then, an important research focus for engineering emerges: how to correlate the production of a company with its logistics chain through intelligent computer systems capable of reading and interpreting input data and sending responses and work orders quickly and accurately.
As a consequence, the logistics system is evolving to a cyber-physical system (CPS) where the cyber-physical system and information system interact with each other, called cyber-physical logistics system (CPLS) (Lai et al. 2013). Cyber-physical Systems assemble the cyber aspects of computing and communications with the dynamics and physics of physical systems operating in the real world (Rajkumar et al., 2012).
Frazzon et al. (2013) introduces and reviews the social aspects of CPS and motivates future research towards Socio-Cyber-Physical Systems (SCPS) applied to production networks. The adoption of Cyber-Physical Systems in production networks enables the new potential for improved efficiency, accountability, sustainability and scalability. In terms of production and transport processes, materializing this potential requires customized technological concepts, planning and control methods as well as the business model.
In terms of production and transport processes, materializing this potential requires customized technological concepts, planning and control methods as well as business models (K. J. Lin and Panahi, 2010).
3.6 DISCUSSION OF THE PAPERS
The main lacks discovered at the literature review were the relationship of VRPPD with simulation and optimization or hybrid approach, and logistic execution systems with technologies CPS.
Just a few papers, e.g. Lai et al (2013), propose optimization model with the focus in minimize the total distribution cost considering uncertainties, and also the vehicle capacity, customer time-window, and the maximum traveling distance as well as the road capacity.
Tan et al (2006) propose a hybrid multi-objective evolutionary algorithm (HMOEA) that incorporates various heuristics for local exploitation in the evolutionary search and the concept of Pareto’s optimality for solving multi-objective optimization in vehicle routing problem with time windows. Some studies involving scheduling with simulation and optimization were also obtained, e.g. the study of Pirard
52
et el (2011) that approach the problem of strategic network design in the context of a multi-site enterprise. They aimed was to present a generic simulation model allowing the managers to evaluate various supply networks designs evaluation and also it takes into account control policies and scheduling of production activities. The simulation model allows carrying out local optimizations by dynamically allocation the customers’ demands and the replenishment orders placed by the sites of the enterprise so as to minimize a cost function. In short, the authors reinforce that the majority of the studied simulation models focuses on the redesign of the distribution network and is developed for a particular industrial case with a specific network structure and thus lack generality models. Rare are the models which study the influence of the availability of production resources, components and raw materials necessary to produce the finished products on the customer’s service performances. Furthermore, most of the developed models do not consider decision-making rules or heuristics which allow us to manage the flows according to the current state of the supply chain.
Lee and Kim (2000) presented a hybrid method combining the analytic and simulation model for an integrated production-distribution system in a supply chain environment. Lin and Chen (2015) present a study of simulation optimization approach for a hybrid flow shop scheduling problem in a real-world semi-conductor back-end assembly facility and the authors suggest that subsequent studies could employ different combinations of other search algorithms and acceleration techniques to improve solution quality while adopting the simulation optimization approach. For the authors, the barriers that hinder the wide application of this approach are the long computation time required for simulation and the noise of performance evaluation using simulation under stochastic conditions. Hernández-Pérez et al. (2016) present a hybrid heuristic approach to solve the multi-commodity pickup-and-delivery traveling salesman problem, which is a routing problem for a capacitated vehicle that has to serve a set of costumers that provide or require a certain amount of m different products, with the objective to minimize the total travel distance.
Gribkovskaia et al. (2007) have proposed a mixed integer programming model and heuristics for the SVRPPD in which each customer may be visited once or twice, giving rise to general solutions that encompass known solution shapes as Hamiltonian, double-path, and lasso. Their results show that the best-known solutions generated by the heuristics are frequently non-Hamiltonian and may contain up to two customers visited twice.
53
Tasan and Gen (2012) contribute to the VRP field by providing an efficient and effective genetic algorithm based approach that introduces highly feasible routes for vehicle routing problem with simultaneous pickups and deliveries (VRP-SPD). In their work, illustrative example and parameter settings were presented followed by the performance evaluation of the proposed approach with computational experiments.
Cyber-physical Systems - CPS assemble the cyber aspects of computing and communications with the dynamics and physics of physical systems operating in the real world (Rajkumar et al., 2012). Frazzon et al. (2013) introduces and reviews the social aspects of CPS and motivates future research towards Socio-Cyber-Physical Systems (SCPS) applied to production networks. The authors reinforce the necessity of future dedicated to the transformation of production networks into networks of Socio-Cyber-Physical Systems. As a necessity, models, measures, and tools for handling aspects related to the embedding of human stakeholders with different individual, organizational and contextual backgrounds need to be developed.
However, the current approaches are limited to scenarios without dynamic influences; i. e. simulation model does not change during the optimization. In fact, transport and logistics scenarios are highly dynamic, so that an appropriate representation of the current system state requires an adaptation of the approach. Therefore, future research involving the application of CPS technologies with methods of simulation-optimization in logistic execution systems constitutes in a research opportunity with potential practical impact.
Wang et al. (2007) developed an approach to evaluated by simulation experiment using dynamic vehicle routing problem with time windows, by an improved evolutionary algorithm to search the best vehicle route in the dynamic network. The authors developed a modified Dijkstra´s algorithm for finding real-time shortest paths using routes attributes and real-time traffic information. The study combined simulation and optimization comparing with IEA, as benchmarks, two famous algorithms: Branch-Bound and Clarke-Wright and the primary conclusion are that the developed approach based on IEA can find best vehicle routes for the DVRPTW efficiently.
Baños et al. (2013) presents an hybrid approach using a new Pareto-based multi-objective approach that uses a multi-start simulated annealing strategy for solving a multi-objective formulation of the VRPTW that aims to minimize the total distance of the vehicles used to service the customers, while also minimizing the imbalance of workloads
