Doi: 10. 1016/S0305-0548(03)00202-8
Download 490.36 Kb. Pdf ko'rish
|
10.1016@S0305-05480300202-8
• (Case1) 1 6 l ¡ L: C = {i ∈ P
i ; l i = l + 1}. P = {i ∈ P i ; l i = l}. • (Case2) l = L, that is, we choose the leaf nodes. Let C = {i ∈ P j : l i = L ∨ v ∈P i x vi = 0}. Let P = P i . • (Case3) l = L + 1: All node pairs. Let C = P = P i . • (Case4) l = L + 2: This case is the same as Case3, but here we skip the checking of the geometrical position of the selected node pairs in the Moves and Swaps procedures (cf. Steps 3 of them). (4) (Local Optimization) Perform Moves(C; P) and then Swaps(C). A.9. Procedure Top-Down-IniSol(N) (1) (Initialize) Let l = 1 denote the actual level of the network to be constructed. Let G denote the set of clusters to be processed and initially let G = {N}. Let H denote a set of clusters and initially let H = ∅. (2) (Loop) In order to assign nodes to level l, ∀G i ∈ G do: (a) (Calculate K min ) Calculate K min depending on the number of nodes in G i , number of ?xed nodes at level l in G i , and the degree and level constraints. (b) (Clustering) Perform Cluster(G i ; K min ; l). Put the resulting clusters to H. If l ¿ 1, connect all new medians M j (found by Cluster in G i ) to the median of cluster G i that was found at the previous level. (3) (Next Level) Let G = H and let H = ∅. Let l = l + 1. If l 6 L, then go to Step 2, otherwise Stop. Top-Down-IniSol applies the following sub-procedures in Step 3 of Cluster: (a) (Allocation) As the exact costs of allocations are not known, the allocation decision is done on the shortest distance basis: ∀i ∈ P /nd the median M j (j = 1; : : : ; K) that is the closest to node i, i.e., d iM j is minimal (see Eq. ( 12 )). Put i to G j , where G j is the jth cluster. (b) (Relocation) Again, as the exact costs of the clusters are not known, the cluster center relocation is done on a geometrical basis. ∀j =1; : : : ; K /nd a new median M j in cluster G j : calculate Geo j denoting the geometrical center of G j , by averaging the (X; Y ) coordinates of the nodes in G j . Then /nd i ∈ G j that is the closest to Geo j , and let new median M j = i. (c) (Cost-calculation) ∀j set M j to be in level l and perform Tree(G j ; M j ; l). If l ¿ 1, then tem- porarily connect all M j to the corresponding median in level l − 1. Finally calculate the cost of the resulted network con/guration according to Eq. ( 1 ) (see Problem de/nition). A.10. Procedure Random-IniSol (N) (1) (Initialize) Select a random node i ∈ N for being at level 1, let the set of connected nodes C = {i}, and let the set of unconnected nodes U = N \ C. Let the multiset of punished nodes C P be empty. (2) (Loop) Select a random j ∈ U. Let the number of trials q = 0. I. G-odor, G. Magyar / Computers & Operations Research 32 (2005) 59–86 85 Table 5 Full-iteration-improvement Advance loop 1 2 3 4 5 6 7 8 9 10 m alloc 1 1 1 1 2 2 2 2 2 2 m locopt 0 1 2 3 1 2 3 3 3 3 ByRoots 1 1 1 1 1 1 1 1 0 0 (3) (Parent Search) While q does not exceed 2 · |C| repeat the following: Select a random parent k ∈ C and let q = q + 1. Try to connect node j to k. If the resulting cost is valid and the connection does not violate any constraint, then perform the connection, take out node k from C if it reached its degree constraint, and go to Step 5. Otherwise, if connecting j to k is not valid, punish node k: put k to C P . If the total number of punishments for node k exceeds a (small) limit, take out k from C. (4) (New Root) If no k satisfying the above conditions was found, let node j be at level 1. (If j is forbidden at level 1, choose another random j ∈ U, which is allowed to be at level 1.) (5) (End) Take out j from U. Put j to C if it is not at the lowest allowed level. If U is not empty, then go to Step 2. A.11. Procedure Full-Iteration-Improvement (N, UseClu, UseLoc) (1) (Initialize) Let C act , C best denote the actual cost, the best cost found, respectively, and let both be initially equal to the cost of the network planned by the initial solution. Let A = 1 assign a column of Table 5 . (2) (Advance Loop) Set the parameters of ReclusterLevel and LocalOpt according to the column A of Table 5 . In the row of parameter ByRoots, 1 denotes TRUE and 0 denotes FALSE. (3) (Level Loop) Sequentially for each level l = 1; : : : ; L do: (a) (ReclusterLevel) If UseClu ∧ l ¡ L, then perform ReclusterLevel(l − 1; m alloc ; m locopt ). (b) (LocalOpt) If UseLoc ∧ A ¿ 7, then perform LocalOpt(l; ByRoots). (4) (Extra LocalOpt1) Calculate C act . If C act ¿ C best (1 − I) ∧ UseLoc, perform LocalOpt(L + 1; ByRoots) and calculate C act . I ∈ (0; 1)1 is a minimal improvement threshold parameter. (5) (Check Cost) If C act ¡ C best (1 − I), let C best = C act and go to Step 3. (6) (Extra LocalOpt2) If A ¿ 9 ∧ UseLoc, perform LocalOpt(L + 2; FALSE). Recalculate C act . (7) (Termination) If A 6 9, let A = A + 1. If C act ¡ C best ∨ A 6 9, let C best = C act and go to Step 2. Otherwise stop, return the actual solution. The parameters are changing according to the scheme presented in Table 5 . References [1] Daskin MS. Network and discrete location. New York: Wiley; 1995. [2] Drezner Z, editor. Facility location. A survey of applications and methods. New York: Springer; 1995. 86 I. G-odor, G. Magyar / Computers & Operations Research 32 (2005) 59–86 [3] Gupta R. Problems in Communication Network Design and Location Planning; New Solution Procedures. PhD thesis, The Ohio State University, 1996. [4] Lo CC, Kershenbaum A. A two-phase algorithm and performance bounds for the star-star concentrator location problem. IEEE Transactions on Communications 1989;37(11):1151–63. [5] Narasimhan S, Pirkul H. Hierarchical concentrator location problem. Computer Communications 1992;15(3):185–91. [6] Pirkul H, Gupta R. Topological design of centralized computer networks. International Transactions in Operational Research 1997;4(1):75–83. [7] Schneider GM, Zastrow MN. An algorithm for the design of multilevel concentrator networks. Computer Networks 1982;6(1):1–11. [8] Soltys JR, Fischer MJ, Roth BD. Optimizing access to service based networks. Telecommunication Systems, Modeling, Analysis, Design and Management 1998;10(3–4):269–90. [9] Klincewicz JG. Hub location in backbone/tributary network design: a review. Location Science 1998;6:307–35. [10] Minoux M. Network synthesis and optimum network design problems: models, solution methods and applications. Network 1989;19:313–60. [11] Brittain D. Optimisation of the Telecommunications Access Network. PhD thesis, University of Bristol, 1999. [12] Anandalingam G, Friesz TL, editors. Hierarchical optimisation. Annals of Operations Research 1992;34:1–11. [13] Eitan Y, Narula SC, Tien JM. A generalized approach to modeling the hierarchical location-allocation problem. IEEE Transactions of Systems, Man, and Cybernetics 1991;21:39–46. [14] Mateus GR, Cruz FRB, Luna HPL. An algorithm for hierarchical network design. Location Science, 1994;2:149–64. [15] Mirchandani PB. Generalized hierarchical facility location. Transportation Science 1987;21:123–25. [16] Serra de la Figuera D, ReVelle CS. The pq-median problem: location and districting of hierarchical facilities. Location Science 1993;1:299–312, 1994;2:63–82. [17] Weaver JR, Church RL. The nested hierarchical median facility location model. INFOR 1991;29:100–15. [18] Harmatos J, JUuttner A, Szentesi (A. Cost-based UMTS transport network topology optimisation. Proceedings of the International Conference on Computer Communications’99, 1999, Tokyo. [19] Harmatos J, Szentesi (A, G(odor I. Planning of Tree-Topology UMTS Terrestrial Access Networks, PIMRC 2000, London, September 2000. [20] Kallenberg P. Optimization of the /xed part of GSM networks using simulated annealing. Proceedings of the Networks’98, 1998, Sorrento, Italy. [21] Kim MJ, Kim SB, Ryu KH. The capacitated hierarchical p-median problem for facility locations of PCS networks. Proceedings of the Fourth International Conference on Telecommunications Systems, 1996, Nashville, TN. [22] Aarts EHL, van Laarhoven PJM. Simulated annealing: theory and applications. Dordrecht: Kluwer Academic Publishers; 1987. [23] Balakrishnan A, Altinkemer K. Using a hop-constrained model to generate alternative communication network designs. INFORMS Journal of Comput. 1992;4:192–205. [24] Gavish B. Topological design of telecommunication networks—local access design methods. Annals of Operations Research 1991;33:17–71. [25] Gavish B. Topological design of telecommunication networks—the overall design problem. European Journal of Operations Research 1992;58:149–72. [26] The UMTS-Forum. http://www.umts-forum.org . [27] Laiho L, Wacker A, Novosad T. Radio network planning and optimisation for UMTS. New York: Wiley; 2002. [28] Jain A, Dubes R. Algorithms for clustering data. Englewoods Cli?s, NJ: Prentice-Hall; 1988. Download 490.36 Kb. Do'stlaringiz bilan baham: |
Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling
ma'muriyatiga murojaat qiling