Mathematical formulations and optimization algorithms for solving rich vehicle routing problems.

Palomo Martínez, Pamela Jocelyn (2018) Mathematical formulations and optimization algorithms for solving rich vehicle routing problems. Doctorado thesis, Universidad Autónoma de Nuevo León.

Vista previa

Texto 1080290389.pdf - Versión Aceptada Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (2MB) | Vista previa

Resumen

Objectives and methods of study: The main objective of this work is to analyze and solve three different rich selective Vehicle Routing Problems (VRPs). The first problem is a bi-objective variant of the well-known Traveling Purchaser Problem (TPP) in which the purchased products are delivered to customers. This variant aims to find a route for which the total cost (transportation plus purchasing costs) and the sum of the customers’s waiting time are simultaneously minimized. A mixed integer bi-objective programming formulation of the problem is presented and tested with CPLEX 12.6 within an ǫ-constraint framework which fails to find non-dominated solutions for instances containing more than 10 nodes. Therefore, a heuristic based on relinked local search and Variable Neighborhood Search (VNS) is proposed to approximate the Pareto front for large instances. The proposed heuristic was tested over a large set of artificial instances of the problem. Computational results over small-sized instances show that the heuristic is competitive with the ǫ-constraint method. Also, computational tests over large-sized instances were carried out in order to study how the characteristics of the instances impact the algorithm performance. The second problem consists of planning a selective delivery schedule of multiple products. The problem is modeled as a multi-product split delivery capacitated team orienteering problem with incomplete services, and soft time windows. The problem is modeled through a mixed integer linear programming formulation and approximated by means of a multi-start Adaptive Large Neighborhood Search (ALNS) metaheuristic. Computational results show that the multi-start metaheuristic reaches better results than its classical implementation in which a single solution is build and then improved. Finally, an Orienteering Problem (OP) with mandatory visits and conflicts, is formulated through five mixed integer linear programming models. The main difference among them lies in the way they handle the subtour elimination constraints. The models were tested over a large set of instances of the problem. Computational experiments reveal that the model which subtour elimination constraints are based on a single-commodity flow formulation allows CPLEX 12.6 to obtain the optimal solution for more instances than the other formulations within a given computation time limit. Contributions: The main contributions of this thesis are: • The introduction of the bi-objective TPP with deliveries since few bi-objective versions of the TPP have been studied in the literature. Furthermore, to the best of our knowledge, there is only one more work that takes into account deliveries in a TPP. • The design and implementation of a hybrid heuristic based on relinked local search and VNS to solve the bi-objective TPP with deliveries. Additionally, we provide guidelines for the application of the heuristic when different characteristics of the instances are observed. • The design and implementation of a multi-start adaptive large neighborhood search to solve a selective delivery schedule problem. • The experimental comparison among different formulations for an OP with mandatory nodes and conflicts.

Actions (login required)

Ver elemento