Evolutionary approaches to o - line routing in backbonecommunications networks

O -line routing in backbone communications networks is an important combinatorial optimisation problem. It has three main uses: rst, o -line routing provides reference benchmark results for dynamic (on-line) routing strategies. Second, and more interestingly, o -line routing is becoming more and more investigated and employed in its own right as a way of quickly nding signi cantly improved routings for live networks which can then be imposed on the network to o er a net improvement in quality of service. Third, it can be used in networks where bandwidth may be booked in advance. In this paper we introduce and investigate a number of heuristic techniques applicable to the routing problem for use in stochastic, iterative search. Results are presented which indicate that these heuristics signi cantly improve the search for solutions, particularly when on-line performance is considered. We also investigate how computation time can be further reduced by the use of delta-evaluation of solutions. Previously, simulated annealing has had a signi cant advantage over genetic algorithms on problems where delta-evaluation is possible because the technique is not usually applicable in genetic algorithms employing standard crossover operators. However, we introduce a specialised crossover operator which enables a genetic algorithm to exploit delta-evaluation e ectively on this problem. We compare the performance of the genetic algorithm employing the new crossover operator with a mutation-only genetic algorithm and nd the performance of the former to be signi cantly better. Nonetheless, results indicate that when our heuristics are used in a simulated annealer and initialisation is performed in a certain way, it outperforms our genetic algorithm on a range of problems of di erent sizes and di culties. This nding leads us to investigate a local search approach to a multiobjective formulation of the routing problem. To this end, a new local search multiobjective optimiser is described, and its performance on the routing problem is compared with that of a multiobjective EA based on the Niched Pareto GA. The ndings from these experiments indicate important roles for both local search and population-based search methods when addressing the multiobjective form of this problem.