C. Gazen and C. Ersoy,“Genetic algorithms for designing multihop lightwave network  topologies,”Artificial Intelligence in Engineering, Elsevier Publishers, Vol.13, Issue3, pp. 211-221,  July 1999.

Multihop lightwave networks are a means of utilizing the large bandwidthof optical fibers. In these networks, each node has a fixed number of transmittersand receivers connected to a common optical medium. A multihop topologyis implemented logically by assigning different wavelengths to pairs oftransmitters and receivers. By using tunable lasers or receivers, it ispossible to modify the topology dynamically when node failures occur ortraffic loads change. The reconfigurability of logical multihop lightwavenetworks requires that optimal topologies and flow assignments be found.In this paper, optimization of these logical topologies by genetic algorithmsis investigated. The genetic algorithm takes topologies as individualsof its population, and tries to find optimal ones by mating, mutating andeliminating them. During the evolution of solutions, minimum hop routingwith flow deviation is used to assign flows, and evaluate the fitness oftopologies. The algorithm is tested with different sets of parameters andtypes of traffic matrices and the solutions are compared against histogramsof random samples from the solution space. These tests show that the solutionsfound by the genetic algorithm are comparable with and in some cases betterthan those found by existing heuristic algorithms.
 

Paper in Adobe Acrobat (310 KBytes)

Back to List of Papers
Back to NETLAB home page
Back to Cem Ersoy'shome page