The coevolutionary supply chain

The coevolutionary supply chain

Process Systems Engineering 2003 B. Chen and A.W. Westerberg (editors) 9 2003 Published by Elsevier Science B.V. 487 The Coevolutionary Supply Chain...

365KB Sizes 3 Downloads 186 Views

Process Systems Engineering 2003 B. Chen and A.W. Westerberg (editors) 9 2003 Published by Elsevier Science B.V.

487

The Coevolutionary Supply Chain Ben H U A a, Jinbiao Y U A N a, David C.W. Hui b China University of Technology, Guangzhou, 510640, P. R. China Tel : +86-20-87113744, Fax" +86-20-85511507, Email:[email protected] b Hong Kong University of Science & Technology, Clear Water Bay, Hong Kong a South

Abstract The new millennium presents many challenges for the enterprises in supply chains. Competitive advantages no longer result from special designs or proper operations of supply chains, but rather the survival of supply chains. This paper attempts to make a biologic digital supply chain paradigm with coevolutionary mechanisms from a new viewpoint. Keywords

supply chain management, supply chain design, coevolution

1. INTRODUCTION In the challenging economy and chaos business, the enterprises are moving focus onto their supply chains. Information technology contributes the most to the presentation and the development of supply chain management. Software vendors present many supply chain management suits. These suits help to manage the whole supply chain rather than to make a good supply chain, i.e. better performance of a supply chain does not come from concession to software. There are two methods to make a good supply chain, one is transform, and the other is evolution. Business Process Reengineering is the representative of transform methods, which often seeks radical redesign and drastic improvement of processes. BPR gurus Michael Hammer and James Champy[1~ note in their book Reengineering the Corporation that only about 30 percent of reengineering projects they have reviewed are successful. One of the primary reasons for this low success rate is that the analyses behind performance estimates are often prepared with flowcharts and spreadsheets. In contrast, evolution methods tend to focus on incremental change and gradual improvement of processes and units of supply chains, and then smooth evolution of supply chains themselves. There are many indicators showing what is a good supply chain. None of the indicators would decides respectively whether a supply chain is good. Because the processes and unites of supply chains always interact to make collective effects on supply chain performance. We attempt to seek the evolution of collective solutions to the problems with coevolution mechanisms. More importantly, coevolution emphasizes interactions of the members; just like that supply chain management emphasizes coordination of the units.

488 2. THE ECOLOGICAL AND BIOLOGICAL SIMILARITY OF SUPPLY CHAINS In the ecosystem, there are individuals, populations, communities, and ecosystem itself at different levels. Correspondingly, we can make such a hierarchy about the markets: individual units (organizations, firms, companies, etc.), so-called suppliers, manufacturers, and purchasers, supply chains, and the whole markets. Ecological and economic systems share many characteristics. [21 Both are complex networks of component parts linked by dynamic processes. Both contain interacting biotic and abiotic components, and are open to exchanges across their boundaries. We reduce the time and space boundaries to ecological communities of ecological systems and supply chains of economic systems. There are still many similarities between supply chains and ecological communities. A supply chain is composed of many units, which may be organizations, companies, or enterprises on the basis of common interests. The units collaborate and synchronously interact with the external markets as a whole supply chain to gain their advantages in the competing markets. The corresponding units in ecological communities are ecological groups. They compose ecological communities based on their survival requirements. In a supply chain, the units are always sorted into supplier, manufacturer, and purchaser. They are connected by flows, such as cost flow, information flow and material flow. E.g. cost flow starts from the purchasers, through the manufacturers, ends at the suppliers. Correspondingly, an ecological group is a species with certain characteristics. This is just like a supply chain unit. These species live harmoniously among a community in an ecosystem. They can also be sorted into hosts and parasites, predators and preys, etc. like the suppliers and the purchasers in a supply chain. They are also connected to each other by flows such as flow of energy, E31flow of information, etc. Fredrik Moberg t4] identified ecological goods and services of coral reef ecosystems, and studied the delivery of ecological goods and services. A supply chain is correspondingly focus on the delivery of goods and services. The need for economics to embrace evolutionary change rather than portray economies as mechanistic systems has long been recognized. [5-6] This means biological and economic systems share many characteristics as ecological and economic systems do. John Foster. [7] studied the economic dynamics based on the biological analogy to economic self-organization. There are many works on communities modeling. E8-91On the other hand, the behavior of social system is similar to that of ecological one in many points, t~ol Every commodity in a market shows time dependent behavior, which is sometimes similar to a life cycle of biological individual. The growth process of a new technology can be understood by the logistic curve, which describes the growth process of a biological species. Industrial structure shows time evolution, which can be related with the change of landscape of ecological community known as the ecological succession. The communities in ecosystems undergo coevolution. (I.e. evolution of two or more interdependent species, each adapting to changes in the other. It occurs, for example, between predators and preys and between insects and the flowers that they pollinate.) Peter T. and Bruce T. Milne [11] described an model of many-species communities to study the principles of ecological community assembly, this enlightened us to study the coevolutionary mechanism of supply chains and

489 the principles which govern a supply chain, such as the competitive advantages of a supply chain corresponding to determining the dominant species in a community. E12-13~.There are other interesting problems; e.g. studying supply chain structure which is similar to an ecological community assembly problem. [141 3. COEVOLUTIONARY COMPUTATION Supply chain management is based on information technology. Computation of coevolution is necessary for studying supply chains from coevolutionary viewpoint. Evolutionary process is used to explore the possibilities inherent in medium it is embedded in. Evolving populations of replicators constantly explore variations around their current forms, without the limitations of preconceptions. GAs and GPs are traditional computational evolution. Their common feature is that the "fitness function" evaluates the genomes in isolation from the other members of population. The genomes interact only with the fitness function (and in some cases the data), but not with each other. This arrangement precludes the evolution of collective solutions to problems, which can be very powerful in supply chain performing. [151 Andr6 L.V. Coelho said: "Artificial coevolution seems very suited for simulating cooperation and/or competition behavior among multiagent entities." in his study on Emergence of multiagent spatial coordination strategies through artificial coevolution. [161 Two basic fl71 classes of coevolutionary algorithms have been developed: competitive coevolution in which the fitness of an individual is determined by a series of competitions with other individuals (see, for example, Rosin and Belew (1996) [181Rosin and R. Belew. New methods for competetive coevolution. Evolutionary Computation, 5(1):1-29, 1996.), and cooperative coevolution in which the fitness of an individual is determined by a series of collaborations with other individuals (see, for example, Potter and De Jong (2000)[19] Potter and K. De Jong. Cooperative coevolution: architecture for evolving coadapted subcomponents. Evolutionary Computation, 8(1): 1-29, 2000. ). Both types of coevolution have been shown to be useful for solving a variety of problems. Evolution proceeds independently, except for evaluation. Since any given individual from a subpopulation represents only a subcomponent of the problem, collaborators will gen = 0

for each species s do Pops(gen) = initialized population evalutate(Pops(gen) )

while not terminated do gen ++

for each species s do Pops(gen)<-select(Pop~(gen- 1)) recombine(Pops(gen) ) evaluate(Pops(gen ) ) survive(Pop~(gen ) )

Fig. 1 The structure of a Cooperative Coevolutionary Algorithm (CCA).

490 need to be selected from the other subpopulations in order to assess fitness. Each generation, all individuals belonging to a particular subpopulation have their fitness evaluated by selecting some set of collaborators from other subpopulations to form complete solutions. Afterward, the CCA proceeds to the next subpopulation, which will in turn draw collaborators from each of the other subpopulations. A simple algorithm of this process is outlined below in figure 1. The competitive coevolutionary algorithm[20-21] is somewhat different from the cooperative like this: in the competitive coevolution, the fitness of an individual in one population is calculated by direct competition with individuals in another population. Individuals of each population must take turns in evaluating and being evaluated for renewing the fitness. By keeping on predominating over each other, some populations will altematively become better than the other. Coevolutionary Algorithm has been used into many fields. [22-24]We attempted to use it into supply chain problems. 4. SUPPLY CHAIN SURVIVAL In ecosystems, Abundance of resources can affect community structure and species abundance, especially through the size of the habitat. Communities can be described as assemblages of interacting populations of species. The structure of a biological community, that is the number (or diversity) and types of species and the number of individuals (or density) in a population is determined by the complex interaction of several factors. Corresponding to the ecological problem, supply chains has the same problem. Managing or understanding the flows between buyers and sellers in supply chains involves design, planning and control of supply chains. On this kind of problem, the beer game is the most famous tool. Here, we suppose another simple beer game to discuss the survival of supply chains. In the survival beer game, it is supposed that there is nothing but supply chain units, i.e. we only discuss the survival of suppliers, manufacturers, and customers in supply chains strategically. "Survival of the fitness" is the basic of Darwin's evolutionary theories. When the thought is used into supply chains, it is interpreted as "Survival of the performance". Supply chain performance is necessary to understand the supply chain and its characteristics. There are many complex supply chain performance measures. We simplify the game's measure as cost, quality and lead-times. Therefore, the survival beer game supposed in this paper is: there are many customers, manufacturers, and suppliers; they interact behind the game, i.e. the performance measures are used to determine the survival of any individual unit among the populations of suppliers, manufacturers, and customers. The outline of the survival beer game is drawn by cooperative coevolution algorithm as Fig. 2. In the Fig. The specifications of the survival beer game are made firstly. The objective that we suppose the survival beer game is to explain the coevolutionary mechanisms of supply chains through cooperative coevolutionary algorithm. In the game, if we initiate the populations with real supply chain units, i.e. suppliers, manufacturers and

491 Gen: the generations of iteration Species: supply chain units, including: supplier, manufacturer and customer Pop: the assembly of species, namely suppliers, manufacturers and customers gen = 0 for each species s do Pops(gen) = initialized population //To initialize it with varies characteristics which will later directly determine the supply //chain performance, i.e. create many varies suppliers, manufacturers and customers evaluate(Pops (gen)) //The evaluation depends on collaborators selected from the other subpopulations in //order to assess fitness of the supply chain performance of the original generation. //(e.g. population of manufacturers selects collaborators from customers and suppliers.) while not terminated do gen ++ for each species s do Pop (gen)<-select(Pops (gen-1)) //To randomly select varies units from the last generation of suppliers, manufacturers //and customers for the new generation. recombine(Pops (gen)) //To form the new populations of current generation.(This implies a new supply chain //constitution, it is better than the that of last generation.) evaluate(Pops (gen)) //To evaluate the supply chain performance of current generation through the //collaborators selected from the other populations. survive(Pops (gen)) //To determine the survival of the units in all populations and consequently let the //supply chain survive the performance.

Fig. 2 The supply chain survive the performance in the survival beer game by Cooperative Coevolutionary Algorithm customers, the game will probably help to determine the selection of supply chain partners. In this way, the game can be particularly used to study smaller ranges of problems, such as evaluating the performance of an individual supply chain unit through using the historical data. This may help a supply chain gain long-term advantages. 5. C O N C L U D I N G R E M A R K S There are many ecological and biological similarities in supply chains. This mades us consider the coevolutionary mechanisms of supply chains. Artificial coevolution provides full supports for the studies of supply chain management problems,

such as the

determination of supply chain structure. The studies of coevolutionary supply chain will promote the progress of academics and practices of supply chain management. ACKNOWLEGEMENT The manuscript benefited greatly from project 79931000 of NSFC and project G2000263 of the State Major Basic Research Development Program. The authors would like to thank Dr. Ming L. Lu. from Aspen Technology, Inc. U.S.A. and Dr, X. Z. Wang from

492 University of Leeds, United Kingdom for giving constructive suggestions. We would also like to extend our thanks to the anonymous referees for their very helpful suggestions.

REFERANCE [ 1] Michael Hammer and James Champy, Reengineering the Corporation: A Manifesto for Business Revolution, 1993. [2] Complex systems and valuation, Ecological Economics 41 (2002) 409-420 [3] Jouni Korhonen a, Margareta Wihersaari, Ilkka Savolainen, Industrial ecosystem in the Finnish forest industry: using the material and energy flow model of a forest ecosystem in a forest industry system, Ecological Economics 39 (2001) 145-161 [4] Ecological goods and services of coral reef ecosystems, Ecological Economics 29 (1999) 215-233 [5] Matthias Ruth, Evolutionary Economics at the Crossroads of Biology and Physics, Journal of Social and Evolutionary Systems Volume: 19, Issue: 2, 1996, pp. 125-144 [6] R.B.Norgaard, The process of loss: exploring the interactions between economic and ecological systems, American Zoologist, 34(1), 1994, pp. 145-158 [7] John Foster, The analytical foundations of evolutionary economics: From biological analogy to economic self-organization, Structural Change and Economic dynamics 8(1997)427-451 [8] Diego J. Rodriguez, A method to detect higher order interactions in ecological communities, Ecological Modelling 117(1999) 81-89 [9] Walter K. Dodds, Geoffrey M. Henebry, Simulation of responses of community structure to species interactions driven by phenotypic change, Ecological Modelling 79(1995) 85-94 [ 10] A Niche Theory of Social Systems, THE 2nd Orwellian Symposium (Karlovy Vary, August, 1994) [ 11 ] Peter T. Hraber, Bruce T. Milne, Community assembly in a model ecosystem, Ecological Modelling 103(1997) 267-285 [ 12] Tilman D., Bildiversity: Poopulation versus ecosystem stability.Ecology 72(2), 350-363,1996 [ 13] Theoretical comparisons of individual success between phenotypically pure and mixed generalist predator populations, Ecological Modelling 82 (1995) 175-191 [ 14] Peter T. Hraber, Bruce T.Milne, Community assembly in a model ecosystem, Ecological Modelling 103(1997) 267-285 [ 15] Paredis J. Coevolutionary computation. Arti.cial Life, Journal 1996;2(4):355-75. [ 16] Emergence of multiagent spatial coordination strategies through artificial coevolution, Andr6 L.V. Coelho, Daniel Weingaertner, Ricardo R. Gudwin, Ivan L.M. Ricarte, Computers & Graphics 25 (2001) 1013-1023. [ 17] R. Paul Wiegand, William C. Liles, Kenneth A. De Jong, An Empirical Analysis of Collaboration Methods in Cooperative Coevolutionary Algorithms, Proceedings of the Genetic and Evolutionary Computation Conference, 2001. [18] Rosin and R. Belew. New methods for competetive coevolution. Evolutionary Computation, 5( 1): 1-29, 1996. [19] Potter and K. De Jong. Cooperative coevolution: An architecture for evolving coadapted subcomponents. Evolutionary Computation, 8(1): 1-29, 2000. [20] Claverie, J.M.; De Jong, K.; Sheta, A.E, Robust nonlinear control design using competitive coevolution, Evolutionary Computation, 2000. Proceedings of the 2000 [21 ] Moeko Nerome, Koji Yamada, Satoshi Endo and Hayao Miyagi, Competitive Co-evolution Based Game-Strategy Acquisition with the Packaging, 1998 Second International Conference on Knowledge-Based Intelligent Electronic Systems, 21-23 April 1998, Adelaide, Australia. Editors, L.C. Jain and R.K, Jain [22] Haoyong Chen and Xifan Wang, Cooperative Coevolutionary Algorithm for Unit Commitment, IEEE Transactions on Power Systems, Volume: 17 Issue: 1, Feb. 2002 Page(s): 128 [23] Qiangfu Zhao, A Co-Eovlutionary Algorithm for Neural Network Learning, International Conference on Neural Networks, Volume: 1,1997 Page(s): 432 -437 vol.1 [24] Weicker, K.; Weicker, N., 1999 Congress on Proceedings of the Evolutionary Computation, 1999 -1632 Vol. 3