Stochastic Capacitated Arc Routing Problem

TL;DR

This paper introduces a genetic algorithm approach to solve the Stochastic Capacitated Arc Routing Problem, focusing on creating robust solutions that handle randomness in arc quantities without significantly increasing costs.

Contribution

It proposes advanced genetic algorithm techniques for SCARP that optimize both cost and robustness, and benchmarks them on standard instances to demonstrate effectiveness.

Findings

Robust solutions can be obtained without significant cost increase.

The method performs well on well-known benchmark instances.

Solutions are applicable to real-world problems with variable quantities.

Abstract

This paper deals with the Stochastic Capacitated Arc Routing Problem (SCARP), obtained by randomizing quantities on the arcs in the CARP. Optimization problems for the SCARP are characterized by decisions that are made without knowing their full consequences. For real-life problems, it is important to create solutions insensitive to variations of the quantities to collect because of the randomness of these quantities. Efficient robust solutions are required to avoid unprofitable costly moves of vehicles to the depot node. Different criteria are proposed to model the SCARP and advanced concepts of a genetic algorithm optimizing both cost and robustness are provided. The method is benchmarked on the well-known instances proposed by DeArmon, Eglese and Belenguer. The results prove it is possible to obtain robust solutions without any significant enlargement of the solution cost. This allows treating more realistic problems including industrial goals and constraints linked to variations in the quantities to be collected.