Co-operation schemes for the parallel memetic algorithm

This paper presents a study of co-operation schemes for the parallel memetic algorithm to solve the vehicle routing problem with time windows. In the parallel co-operative search algorithms the processes communicate to exchange the up-to-date solutions, which may guide the search and improve the results. The interactions between processes are defined by the content of the exchanged data, timing, connectivity and mode. We show how co-operation schemes influence the search convergence and solutions quality. The quality of a solution is defined as its proximity to the best, currently-known one. We present the experimental study for the well-known Gehring and Homberger's benchmark. The new world's best solutions obtained in the study confirm that the co-operation scheme has a strong impact on the quality of final solutions.