Achieving robustness in the capacitated vehicle routing problem with stochastic demands

Abstract

ABSTRACT Stochastic demands can impact the quality and feasibility of a solution. Robust solutions then become paramount. One way to achieve robustness in the Capacitated Vehicle Routing Problem with Stochastic Demands (CVRPSD) is to add a measure of the second-stage (recourse) distance to the objective function of the deterministic problem. We adopt variance as a measure of the recourse distance and propose a Mean-Variance (MV) model. To solve the model, a Hybrid Sampling-based solution approach is developed. Numerical experiments are conducted on benchmark instances and a selective waste collection system in Brazil. We compare our model with others from literature which also use a measure of the second-stage distance to attain robustness. The numerical results show that our model generates the most robust solutions. The comparison provides detailed features of each model and their advantages and disadvantages, helping decision-makers decide which model to utilize based on their different needs and priorities.