A self-adaptive parallel robust tabu-search algorithm for the quadratic assignment problem

With this study, we propose a novel parallel robust tabu-search algorithm (parallel-tabu-qap) for the np-hard quadratic assignment problem (qap)of whose instances having large number of location and facilites have not been reported to be solved exactly so far. parallel-tabu-qap algorithm that has two phases optimizes the parameters of tabu-search algorithm in its first phase by using genetic algorithms through generations. the individuals that have parameters of the tabu-search are optimized in a population that is located on a master processor. in the second phase, slave processors optimize the solution of the problem by restarting their search process in case of stagnations. with its stagnation prevention and parallel optimization talents, parallel-tabu- qap algorithm is observed to obtain better results than its sequential and statically parameter- tuned counterparts. the algorithm has 0.05% deviation for the benchmark tests performed on more than 100 problem instances. these experimental results show that the parallel-tabu-qap algoritm is one the best performing techniques in its heuristics algorithms class when compared with state-of-the-art qap algorithms.