A Study Of RWA Algorithm For Optical WDM Networks Based On Improved Genetic Algorithm

As the key solution to the backbone network in future, WDM (Wavelength Division Multiplexing) optical network is continuously focused and it’s quite reasonable that study on it has been a hot topic recently. Although WDM network has multiplied the bandwidth greatly, available resources, such as fibers and wavelengths, seem still not so adequate in the face of the astonishingly increasing WDM bearer services. Therefore, it becomes definitely important to optimize the routing and wavelength assignment for the WDM network so that some indicators such as resource utilization and blocking probability can be as better as possible, which is right the called RWA problem.After giving a review of current RWA approaches in Chapter2, we propose a new RWA algorithm which is named Double Crossover and Mutation Adaptive-Genetic Algorithm (DCMA-GA) for WDM network based on improved GA in Chapter3. In DCMA-GA, the chromosome is designed as variable-lengthed and natural-numbered in the form of wavelength No. followed by node No. sequences of the lightpath. The initial population is generated by lightpaths which are searched randomly. The fitness function considers both the cost of the lightpath and the influence of the WA on the whole network—we use RLI WA algorithm to calculate this influence. For the selection operation, we use roulette wheel selection combined with best-reserving selection. The crossover and mutation operations are designed as double crossover and mutation—lighthpath first and then wavelength, with independent probabilities respectively. The crossover and mutation probabilities are designed as adaptive.To simulate DCMA-GA and its contrastive algorithms, we develop a simple simulation program and give a brief introduction to it in Chapter4. This simulation program consists of two modules, request generating module and request processing module. The former focuses on generating random requests yielding to Poisson distribution while the latter mainly processes those requests via calling DCMA-GA or other RWA algorithms to set up lightpaths or discard requests once RWA fails.Finally, we simulate the new algorithm, the traditional Dijkstra+FF algorithm and the original GA in the fourteen-noded NSFNET topology with our simulation program. And the statistical results both demonstrate the performance of the new algorithm.