A multi-objective Monte Carlo tree search for forest harvest scheduling

• Tipo: Artigo em Revista Científica Internacional
• Data: 2020

Autores

Neto, T (Autor) Outra; Constantino, M (Autor) Outra; Martins, I (Autor) Outra; Joao Pedro Pedroso (Autor) FCUP

Revista

Título: European Journal of Operational Research

Vol. 282

Páginas: 1115-1126

ISSN: 0377-2217

Editora: Elsevier

Indexação

ISI Web of Knowledge - 6 Citações

Scopus - 7 Citações

Outras Informações

ID Authenticus: P-00R-3FB

DOI: 10.1016/j.ejor.2019.09.034

Abstract (EN): While the objectives of forest management vary widely and include the protection of resources in protected forests and nature reserves, the primary objective has often been the production of wood products. However, even in this case, forests play a key role in the conservation of living resources. Constraining the areas of clearcuts contributes to this conservation, but if it is too restrictive, a dispersion of small clearcuts across the forest might occur, and forest fragmentation might be a serious ecological problem. Forest fragmentation leads to habitat loss, not only because the forest area is reduced, but also because the core area of the habitats and the connectivity between them decreases. This study presents a Monte Carlo tree search method to solve a bi-objective harvest scheduling problem with constraints on the clearcut area, total habitat area and total core area inside habitats. The two objectives are the maximization of both the net present value and the probability of connectivity index. The method is presented as an approach to assist the decision maker in estimating efficient alternative solutions and the corresponding trade-offs. This approach was tested with instances for forests ranging from some dozens to over a thousand stands and temporal horizons from three to eight periods. In general, multi-objective Monte Carlo tree search was able to find several efficient alternative solutions in a reasonable time, even for medium and large instances.

Idioma: Inglês

Tipo (Avaliação Docente): Científica

Nº de páginas: 12