A thin-provisioned and functionalized memetic algorithm for the single row facility layout problem

Abstract

The objective of single row facility layout problem (SRFLP) is to assemble a full permutation of facilities along a straight line so that the weighted sum of pairwise distances is minimized. This category of layout positioning problem was proven to be NP-hard. In this work, a thin-provisioned and functionalized memetic algorithm (TPFMA) for solving this problem is developed. The three main contributions of TPFMA are as follows: (i) a fixed population of four individuals equipped with different roles helps to remove redundancy and maintain the balance between exploitation and exploration; (ii) a pairwise precedences-based crossover contributes to the rapid identification of common good genes; and (iii) a specific probabilistic model assists in repairing infeasible solutions and restarting the population in a low-cost, efficient way. In particular, evaluations over 43 small instances show that this method successfully achieves the optimum solution but employs considerably less computing time than its competitors over quite a few examples. Moreover, TPFMA improves 21 previous best results out of the 40 most challenging large instances when running for the same amount of computing time. The results on other SRFLP instances with up to 1000 facilities suggest that TPFMA matches the best known values for the 40 benchmark instances out of the total 50 instances.