Inserting knowledge in multirecombined evolutionary algorithms for the flow shop scheduling problem

Abstract

Determining an optimal schedule to minimize the completion time of the last job abandoning the system (makespan) becomes a very difficult problem when there are more than two machines in the flow shop. Due both to its economical impact and complexity, different techniques to solve the Flow Shop Scheduling problem (FSSP) has been developed. Current trends addressed to multire-combination, involve distinct evolutionary computation approaches providing not a single but a set of acceptable alternative solutions, which are created by intensive exploitation of multiple solutions previously found. Evolutionary algorithms perform their search based only in the relative fitness of each potential solution to the problem. On the other hand specialised heuristics are based on some specific features of the problem. This work shows alternative ways to insert knowledge in the search by means of the inherent infor-mation carried by solutions coming from that specialised heuristic or gathered by the evolutionary process itself. The present paper compares the performance of multirecombined evolutionary algo-rithms with and without knowledge insertion and their influence in the crossover rate, the popula-tion size and the quality of results when applied to selected instances of the FSSP.