Genetic algorithms for the fixed charge transportation problem

摘要

Two genetic algorithms (GAs) are presented for the fixed charge transportation problem (FCTP). Both algorithms incorporate knowledge about the properties of optimal solutions. The algorithms mainly differ in the technique used to deal with the inherent constraints of FCTP: while the GA based on the permutation representation can be classified as a decoder approach, the matrix representation resembles direct solution encoding with specialized operators to maintain feasibility. The authors compare both GAs on randomly generated instances. The simulations indicate the suitability of the approaches-in particular of the matrix representation-and demonstrate their superiority to a GA which has been proposed for transportation problems.