Models and algorithms for berth allocation problems in port terminals

Resumen

Seaports play a key role in maritime commerce and the global market economy. Goods of different kinds are carried in specialized vessels whose handling requires ad hoc port facilities. Port terminals comprise the quays, infrastructures, and services dedicated to handling the inbound and outbound cargo carried on vessels. Increasing seaborne trade and ever-greater competition between port terminals to attract more traffic have prompted new studies aimed at improving their quality of service while reducing costs. Most terminals implement operational planning to achieve more efficient usage of resources, and this poses new combinatorial optimization problems which have attracted increasing attention from the Operations Research community. One of the most important problems confronted at the quayside is the efficient allocation of quay space to the vessels calling at the terminal over time, also known as the Berth Allocation Problem. A closely related problem arising in terminals that specialize in container handling concerns the efficient assignment of quay cranes to vessels, which, together with quay space planning, leads to the Berth Allocation and Quay Crane Assignment Problem. These problems are known to be especially hard to solve, and therefore require designing methods capable of attaining good solutions in reasonable computation times. This thesis studies different variants of these problems considering well-known and new real-world aspects, such as terminals with multiple quays or irregular layouts. Mathematical programming and metaheuristics techniques are extensively used to devise tailored solution methods. In particular, new integer linear models and heuristic algorithms are developed to deal with problem instances of a broad range of sizes representing real situations. These methods are evaluated and compared with other state-of-the-art proposals through various computational experiments on different benchmark sets of instances. The results obtained show that the integer models proposed lead to optimal solutions on small instances in short computation times, while the heuristic algorithms obtain good solutions to both small and large instances. Therefore, this study proves to be an effective contribution to the efforts aimed at improving port efficiency and provides useful insights to better tackle similar combinatorial optimization problems.