Double archive Pareto local search

Many real-world problems have multiple, conflicting objectives and a large complex solution space. The conflicting objectives give rise to a set of non-dominating solutions, known as the Pareto front. In the absence of any prior information on the relative importance of the objectives, none of these solutions can be said to be better than others, and they should all be presented to the decision maker as alternatives. In most cases, the number of Pareto solutions can be huge and we would like to provide a good representative approximation of the Pareto front. Moreover, the search space can be too large and complex for the problem to be solved by exact methods. Therefore efficient heuristic search algorithms are needed that can handle such problems. In this paper, we propose a double archive based Pareto local search. The two archives of our algorithm are used to maintain (i) the current set of non-dominated solutions, and (ii) the set of promising candidate solutions whose neighbors have not been explored yet. Our selection criteria is based on choosing the candidate solutions from the second archive. This method improves upon the existing Pareto local search and queued Pareto local search methods for bi-objective and tri-objective quadratic assignment problem.