A hybrid genetic algorithm for the provision of isochronous service in high speed networks

Abstract

A hybrid genetic algorithm, which combines the concepts of genetic search and hillclimbing is proposed in order to solve the problem of Isochronous Channel Reusing (ICR) in the topologies of LANs/MANs. The problem of ICR, i.e. the task to assign channels to newcoming requests with the twofold issue of obtaining the least number of employed channels while no overlapped isochronous requests/connections share the same isochronous channel(s), is an NP complete optimization problem (Nen-Fu Huang and Huey-Ing Liu, 1994). The special features of the proposed algorithm, which we call Hybrid Genetic Reuse Algorithm (HGRA), that distinguish HGRA from the Simple Genetic Algorithm (SGA) (D.E. Goldber, 89) are: (1) a binary two dimensional encoding; (2) the use of a more advanced fitness function in the phase of selection; (3) a repair strategy until a feasible solution is produced; and (4) a local improvement operator. The computational complexity is also computed. The proposed algorithm was tested through simulation on a dual bus Metropolitan Area Network (DQDB) and was compared to a graph coloring algorithm named ICRA (Nen-Fu Huang and Huey-Ing Liu, 1994), presented on a DQDB network. The experimental results indicate that a genetic algorithm, with its properties: robustness, implicit parallelism, combination and exploration capabilities hybridized with a strong local search algorithm, can be used for finding good solutions for the problem of ICR, with good scalability property, comparable to those of ICRA and even better in case of heavy network load.