Iterated greedy with variable neighborhood search for a multiobjective waste collection problem. (1st May 2020)

• Record Type: Journal Article
• Authors: Delgado-Antequera, Laura; Caballero, Rafael; Sánchez-Oro, Jesús; Colmenar, J Manuel; Martí, Rafael
• Is Part Of: Expert systems with applications. Volume 145(2020)
• Publication Date: 2020-05-01
• Subjects: Metaheuristics -- Multiobjective optimization -- Vehicle routing; Expert systems (Computer science) -- Periodicals; Systèmes experts (Informatique) -- Périodiques; Electronic journals; 006.33
• DOI: 10.1016/j.eswa.2019.113101
• Languages: English
• ISSNs: 0957-4174

Abstract

Highlights: Considering a real routing problem and modeling it with 4 objective functions. Solving system combining metaheuristics with achievement scalarizing function. Developing the methodology to apply iterated greedy to a multi objective problem. Performing exhaustive experimentation to test the method. Improving both the currently implemented solution and NSGA-II. Abstract: In the last few years, the application of decision making to logistic problems has become crucial for public and private organizations. Efficient decisions clearly contribute to improve operational aspects such as cost reduction or service improvement. The particular case of waste collection service considered in this paper involves a set of economic, labor and environmental issues that translate into difficult operational problems. They pose a challenge to nowadays optimization technologies since they have multiple constraints and multiple objectives that may be in conflict. We therefore need to resort to multiobjective approaches to model and solve this problem, providing efficient solutions in short computational times. In particular, we consider four different objectives to model the waste collection problem: travel cost, route length balance, route time balance, and number of routes. We propose an iterated greedy algorithm coupled with a variable neighborhood search to minimize an achievement function to determine a good approximation to the Pareto front. The performance of our method is empirically analyzed on a set of instances (both generated and real), and compared with the well-known NSGA-II and SPEA2 methods. The comparison favors our proposal.