Strictly nonblocking grooming of dynamic traffic in unidirectional SONET/WDM rings using genetic algorithms

We propose a genetic algorithm (GA) to solve the strictly nonblocking grooming problems with dynamic traffic in unidirectional synchronous optical network/wavelength division multiplexing rings with arbitrary asymmetric traffic patterns. We give out the theoretical lower and upper bounds on the numbers of both add/drop multiplexers (ADMs) and wavelengths required and define the blocking properties in grooming dynamic traffic in detail. In applying GAs to this problem, we propose a first fit approach incorporated with a greedy local improvement algorithm to decode the chromosome. Four different local improvement scenarios are designed and compared in this paper. We also introduce the concepts of Link Load Imparity, Weighted Traffic Imparity and Average Link Load of each traffic pattern and use the latter two to estimate the first parameter before grooming. Comparison between the genetic and the random search results is also made. Computer simulation results show that it is important to combine the GA with an appropriate local improvement scenario in order to achieve near optimal results and to reduce the running time and that the algorithm proposed in this paper is able to achieve good results in reducing the numbers of ADMs and wavelengths. (C) 2002 Elsevier Science B.V. All rights reserved.