Optimizing Tree Selection for Sustainable Logging in Amazonian Areas with Species Restrictions Using Bio-Inspired Algorithms

Forest management planning in the Peruvian Amazon must reconcile economic return, ecological sustainability, spatial coherence, and strict regulatory compliance within a single harvest decision. Biomimetic optimization—search that transposes natural mechanisms such as honeybee foraging, biological evolution, and physical annealing into computation—offers a promising route to such constrained combinatorial problems. This study proposes a set-theoretic discrete adaptation of the Artificial Bee Colony algorithm (Set-ABC), a swarm-intelligence metaheuristic inspired by the division of labor among employed, onlooker, and scout bees, for parcel-based tree selection; food sources are represented as sets of parcel identifiers and neighborhood search is driven by set-algebraic operators rather than continuous or binary encodings. The task is formalized as a regulated NP-hard combinatorial optimization with hard constraints anchored in Law No. 29763 and Supreme Decree No. 018-2015-MINAGRI, plus a penalized species-retention constraint, under an objective maximizing net economic return in monetary units (PEN). Set-ABC is benchmarked against three further nature-inspired paradigms—a Binary ABC (BABC), a Binary Genetic Algorithm (GA), and Simulated Annealing (SA)—on a semi-synthetic instance of approximately 90 parcels and 12 commercial species from a real operational inventory. After Grid Search calibration, the methods were evaluated over 20 paired runs. A Friedman test with Bonferroni-corrected Wilcoxon post hoc comparisons revealed significant differences across all pairs: GA attained the highest mean net return (1,059,364 PEN) and Set-ABC the lowest (983,873 PEN) but the second-most stable variance, establishing this biomimetic set-based formulation as a viable, stable contender for regulated parcel-based forest optimization.