The Characteristic of Logistics Network Specific Coding in Genetic Algorithms

Abstract

Logistics network related problems are usually associated with geographical locations, but most of evolutionary computing heuristics such as genetic algorithms (GA) in solving them have not given appropriate labeling for locations. By our experiments, it can introduce fatal failure sometimes. However, once we took the linkage information into consideration, we found the linkage learning genetic algorithm (LLGA) is a more stable solution method possessing the independence of coding schemes for facility location problems than the simple genetic algorithm. Except for what mentioned above and many parameter-setting experiments, we also address the way to improve the performance of the LLGA with more but limited interpretation points. All of these can be good guidelines for users interested in applying the evolutionary computing to solve logistics network related problems.