Analysis of The Ant Number Effects on Ant Colony Optimization for Solving Russia-20-Nodes-SDVRP Instance

Abstract

SDVRP is a variation of the CVRP that allows different vehicles to visit the same node. ACO performance is influenced by it’s parameter value. The number of ants parameter has a strong influence on ACO Performance, because it influences directly on solution quality and computation time. It is too hard to find the parameter value that can make ACO performance being the best performance. Because of this, what can be done is to test several parameter values, then compare the ACO performance of each parameter value. Russia-20-Nodes-SDVRP Instance is attractive to be solved because it’s number of nodes are not too many, only 20 nodes. Based on that number of nodes, the computation time of this instance is relatively fast and it has a lot of solution candidate. This research focus on the ant number effects on ACO for Solving Russia-20-Nodes-SDVRP Instance. Five ant numbers will be tested, they are 5, 10, 15, 20 and 25 ants, each ant number will be tested 10 times, so there will be 50 tests. These tests show that more ants used, greater possibility of getting a better route, but longer calculation time needed. The decrease in average of length of route is 1,1521% while the increase in average of calculation time is 42.5525%.