Enhanced Parallel Tabu Search with the Memory of Local Optima

Abstract

Tabu search is a widely used heuristic search method. However, there are two drawbacks in traditional tabu search. First, tabu search, as well as other search methods, only provides the best solution obtained during the search process, and there is no way to know the quality of the obtained solutions. Second, tabu list helps tabu search avoid the problem of looping in a small cycle, but it cannot prevent tabu search from searching previously searched areas again or, worse, looping in a large cycle. The computation time of a search method can be reduced by implementing parallel processing. This study proposes a parallel deterministic simple tabu search, which computes more efficiently and overcomes the two drawbacks mentioned above. The results of our experiments show that it takes less computation effort for the proposed parallel tabu search to find a global optimal solution than for a conventional parallel tabu search.