Dynamic tabu list management using the reverse elimination method

Tabu search is a metastrategy for guiding known heuristics to overcome local optimality. Successful applications of this kind of metaheuristic to a great variety of problems have been reported in the literature. However, up to now mainly static tabu list management ideas have been applied. In this paper we describe a dynamic strategy, the reverse elimination method, and give directions on improving its computational effort. The impact of the method will be shown with respect to a multiconstraint version of the zero-one knapsack problem. Numerical results are presented comparing it with a simulated annealing approach.