A Communication Scheme for an Experimental Grid in the Resolution of VRPTW Using an Evolutionary Algorithm

In this work is defined a scheme of two stages for sending population segments of one Parallel Genetic Algorithm (PGA) to the nodes of an experimental grid called "Tarántula miniGrid". The technique used to link the clusters and to configure the Tarántula mini Grid is described and the effects of latency in the communications between these clusters and the behavior of the speedup are discussed. The PGA is used for solving instances of the Vehicle Routing Problem with Time Windows (VRPTW). Results of this approach show that this communication scheme increases the efficiency of PGA in a grid environment and produces greater exploration and exploitation of the solution space of VRPTW.