A vehicle routing improvement algorithm comparison of a "greedy" and a matching implementation for inventory routing

Given an initial solution to a vehicle routing problem-(VRP)-we present three heuristic route improvement schemes based on the concept of node interchange between different routes. Three algorithms are presented together with their computational performance when applied to an inventory routing problem for 12 consecutive weekly periods. All three procedures improve the initial solution constructed by a modified Clark and Wright algorithm by over 50% as measured by the change in the value of the objective function. In two of the algorithms a minimum weight matching problem is solved repetitively.