物流管理英文-含翻译.doc

An internet-based logistics management system for enterprise chains
N。 Prindezis， C.T。 Kiranoudis
School of Chemical Engineering, National Technical University， 15780 Athens, Greece
Received 13 September 2003; received in revised form 20 December 2003； accepted 27 January 2004
Available online 10 December 2004
Abstract
This paper presents an Internet—Based Logistics Management System to coordinate and disseminate tasks and related information for solving the heterogeneous vehicle routing problem using appropriate metaheuristic techniques， for use in enterprise chain net works. Its architecture involves a JAVA Web applet equipped with interactive communication capabilities between peripheral software tools。 The system was developed in distributed software fashion technology for all computer platforms utilizing a Webbrowser， focusing on the detailed road network of Athens and the needs of the Athens Central Food Market enterprises. 2004 Elsevier Ltd。 All rights reserved。
Keywords: Decision support system; e-Logistics; Transportation； Vehicle routing problem
1. Introduction
Enterprise chains are the business model of the present and future regarding markets that involve small and medium company sizes。 Clearly, grouping activities towards a focused target facilitates an understandably improved market penetration guaranteed by a successful trade mark of a leading company in the field。 Several collaboration models that basically include franchising are introduced as a part of this integrated process. When such a network is introduced in order to exploit a commercial idea or business initiative and subsequently expanded as market penetration grows， several management issues arise regarding the operations of the entire network. Such a network is the ideal place for organizing and evaluating in a more centralized way several ordinary operations regarding supply chain and logistics Infact， tools developed for organizing management processes and operational needs of each individual company， can be developed in a more centralized fashion and the services provided by the tool can be offered to each network member to facilitate transactions and tackle operations similarly. Web-based applications are an ideal starting place for developing such applications. Typically such systems serve as a central depot for distributing common services in the field of logistics。 The commercial application is stored in a central server and services are provided for each member of the group。 A prototype of such a server is described in a previous work (Prindezis， Kiranoudis, & Marinos—Kouris,2003）。 This paper presents the completed inter net system that is installed in the central web server of the Athens Central Food Market that deals with the integrated problem of distribution for 690 companies that comprise a unique logistics and retail chain of enterprises. The needs of each company are underlined and the algorithms developed are described within the unified internet environment. The problem solved and services provided for each company is the one involving distribution of goods through a heterogeneous fleet of trucks. New insights of the metaheuristics employed are provided. A characteristic case study is presented to illustrate the effectiveness of the proposed approach for a real—world problem of distribution through the detailed road network of Athens。
2. Distribution through heterogeneous vehicle fleets
The fleet management problem presented in this paper requires the use of a heterogeneous fleet of vehicles that distribute goods through a network of clients
(Tarantilis， Kiranoudis, ＆ Vassiliadis, 2003, 2004).Therefore， the system was designed in order to automatically generate vehicle routes （which vehicles should de—
liver to which customers and in which order), using rational， quantitative, spatial and non-spatial information and minimizing simultaneously the vehicle cost and the total distance travelled by the vehicles， subject to the following constraints:
l each vehicle has a predetermined load capacity， typically different from all other vehicles comprising the fleet （heterogeneous nature),
l the capacity of a vehicle cannot be exceeded,
l a single vehicle supplies each customers demand,
l the number of vehicles used is predetermined.
The problem has an obvious commercial value and has drawn the attention of OR community。 Its great success can be attributed to the fact that it is a very interesting problem both from the practical and theoretical points of view. Regarding the practical point of view， the distribution problem involved definitely plays a central role in the efficiency of the operational planning level of distribution management， producing economical routes that contribute to the reduction of distribution costs, offering simultaneously significant savings in all related expenses (capital, fuel costs, driver salaries）。 Its Importance in the practical level, motivated in tense theoretical work and the development of efficient algorithms。
For the problem by academic researchers and professional societies in OR/MS, resulting in a number of papers concerning the development of a number of Vehicle Routing Information Systems （VRIS） for solving the problem。 The problem discussed is an NP—hard optimization problem, that is to say the global optimum of the problem can only be revealed through an algorithm of exponential time or space complexity with respect to problem size。 Problems of this type are dealt with heuristic or metaheuristic techniques。 Research on the development of heuristic algorithms （Tarantilis ＆ Kiranoudis, 2001，2002a， 2002b) for the fleet management problem has made considerable progress since the first algorithms that were proposed in the early 60s。 Among them， tabu search is the champion （Laporte， Gendreau， Potvin， ＆ Semet,2000)。 The most powerful tabu search algorithmsare now capable of solving medium size and even largesize instances within extremely small computational environments regarding load and time. On the algorithmic side， time has probably come to concentrate on the development of faster， simpler （with few parameters） and more robust algorithms, even if this causes a small loss in quality solution。 These attributes are essential if an algorithm is to be implemented in a commercial package。
The algorithm beyond the system developed is of tabu search nature。 As mentioned before, since the algorithms cannot reveal the guaranteed global optimum, the time that an algorithm is left to propose a solution to the problem is of utmost importance to the problem。 Certainly， there is a trade—off between time expected for the induction of the solution and its quality. This part was implemented in a straightforward way. If the system is asked by the user to produce a solution of very high quality instantly, then an aggressive strategy is to be implemented。 If the user relaxes the time of solution to be obtained， that is to say if the algorithm is left to search the solution space more effciently， then there is room for more elaborate algorithms.
The algorithm employed has two distinct parts. The first one is a generalized route construction algorithm that creates routes of very good quality to be improved by the subsequent tabu phase. The construction algorithm takes into account the peculiarities of the heterogeneous nature of fleet and the desire of the user to use vehicles of his own desire， owned or hired， according to his daily needs.
The Generalized Route Construction Algorithm employed, is a two—phase algorithm where unrouted customers are inserted into already constructed partial solutions. The set of partial solutions is initially empty， and in this case a seed route is inserted that contains only the depot。 Rival nodes to be inserted are then examined.
All routes employed involve single unrouted customers. The insertion procedure utilizes two criteria c1(i，u,j) and c2(i，u，j） to insert a new customer u between two adjacent customers i and j of a current partial route. The first criterion finds the best feasible insertion point (i ＊,j *) that minimizes the Clark and Wright saving calculation for inserting a node within this specific insertion point，
C1(i,u,j)=d(I，u）+d（u,j）—d（I，j） (1)
In this formula， the expression d（k，l） stands for the actual cost involved in covering the distance between nodes k and l。 The Clark and Wright saving calculation introduced in this phase serves as an appropriate strong intensification technique for producing initial constructions of extremely good quality， a component of utmost
necessity in tabu improvement procedure。
The second phase involves the identification of the actual best node to be inserted between the adjacent nodepair （i＊ ，j *) found in the first phase (Solomon, 1987). From all rival nodes， the one selected is the one that maximizes the expression
C2 （i*， u, j ＊)=[d(0，u）+d（u，0）]— C1(i＊， u， j *) （2）
where 0 denotes the depot node. The expression selected is the travelling distance directly from/to the depot to/ from the customer and the additional distance expressed
by the first criterion。 In all, the first phase of the construction algorithm seeks for the best insertion point in all possible route seeds and when this is detected, the appropriate node is inserted. If no feasible node is found， a new seed route， containing a single depot, is inserted.
The algorithm iterates until there are no unrouted nodes. It must be stretched that the way routes are filled up with customers is guided by the desire of the user regarding
the utilization of his fleet vehicles。 That is to say, vehicles are sorted according to the distribution and utilization needs of the dispatcher. Vehicles to be used first （regarding to user cost aspects and vehicle availability） will be loaded before others that are of lower importance to the user. Typically， all users interviewed expressed the desire for the utilization of greater tonnage vehicles instead of lower tonnage， so vehicles for loading were sorted in descending order of capacity.
For the subsequent aggressive part of the algorithm a tabu search metaheuristic was implemented。 The basic components of this algorithm employed in this application are the neighbourhood definition， the short-term memory and the aspiration criterion.
2。1. Neighbourhood
The neighbourhood is defined as a blend of the most favorable local search moves that transforms one solution to another。 In particular, in its tabu search iteration the type of move adopted is decided stochastically. A predefined probability level is assigned to each move type. After that, it is decided whether the move operation is performed within a single route or between different routes, once more stochastically。 This time, for both operations, the probability level is assigned a value of 50％。 Subsequently， the best neighbour that the selected move implies is computed。 The move types employed are the 2-Opt move (Bell et al。， 1983)， the 1–1 Exchange move (Evans& Norback ， 1985）， the1–0 Exchangemove （Evans ＆ Norback， 1985）, on both single route and different routes.
2.2. Short—term memory
Short—term memory， known as tabu list, is the most often used component of tabu search. Tabu list is imposed to restrict the search from revisiting solutions that were considered previously and to discourage the search process from cycling between subsets of solutions。 For achieving this goal， attributes of moves, more precisely the reversals of the original ones， are stored in a tabu list。 The reversal moves that contain attributes stored in tabu list are designated tabu and they are excluded from the search process。 Regarding the tabu search variant implemented， these attributes are the nodes involved in the move （all the moves used in the this method can be characterized by indicating only two nodes) and the corresponding routes where these nodes belong to。 The number ofiterations that arcs’mobility is restricted is known as tabu list size or tabu tenure. The management of the tabu list is achieved by removing the move which has been on the tabu list longest。
2.3。 Aspiration criterion
The aspiration criterion is a strategy for overriding the short-term memory functions。 The tabu search method implemented, uses the standard aspiration criterion: if a move gives a higher quality solution than the best found so far, then the move is selected regardless its tabu status.
Tabu Search algorithm terminates when the number of iterations conducted is larger than the maximum number of iterations allowed。
3。 Developing the internet-based application tool
Web services offer new opportunities in business landscape， facilitating a global marketplace where business rapidly create innovative products and serve customers better. Whatever that business needs is, Web services have the flexibility to meet the demand and allow to accelerate outsourcing. In turn， the developer can focus on building core competencies to create customer and shareholder value。 Application development is also more efficient because existing Web services， regardless of where they were developed， can easily be reused。
Many of the technology requirements for Web services exist today, such as open standards for business to—business applications， mission-critical transaction platforms and secure integration and messaging products. However， to enable robust and dynamic integration of applications， the industry standards and tools that extend the capabilities of to days business—to-business interoperability are required。 The key to taking full advantage of Web services is to understand what Web services are and how the market is likely to evolve. One needs to be able to invest in platforms and applications today that will enable the developer to quickly and effectively realize these benefits as well as to be able to meet the specific needs and increase business productivity.
Typically， there are two basic technologies to be implemented when dealing with internet—based applications; namely server-based and client-based。 Both technologies
have their strong points regarding development of the code and the facilities they provide。 Server-based applications involve the development of dynamically created web pages。 These pages are transmitted to the web browser of the client and contain code in the form of HTML and JAVASCRIPT language。 The HTML part is the static part of the page that contains forms and controls for user needs and the JAVASCRIPT part is the dynamic part of the page。 Typically, the structure of the code can be completely changed through the intervention of web server mechanisms added on the
transmission part and implemented by server—based languages such as ASP， JSP， PHP, etc。 This comes to the development of an integrated dynamic page application where user desire regarding problem peculiarities （calculating shortest paths, execute routing algorithms, transact with the database， etc。）