A general variable neighborhood search for solving the multi-objective open vehicle routing problem

Abstract

The multi-objective open vehicle routing problem (MO-OVRP) is a variant of the classic vehicle routing problem in which routes are not required to return to the depot after completing their service and where more than one objective is optimized. This work is intended to solve a more realistic and general version of the problem by considering three different objective functions. MO-OVRP seeks solutions that minimize the total number of routes, the total travel cost, and the longest route. For this purpose, we present a general variable neighborhood search algorithm to approximate the efficient set. The performance of the proposal is supported by an extensive computational experimentation which includes the comparison with the well-known multi-objective genetic algorithm NSGA-II.