Application of relax-and-fix heuristic in the aggregation of stands for tactical forest scheduling

Abstract Tactical harvest scheduling is related to forest unit selection and spatial sequencing. Mixed Integer Linear Programming (MILP) is traditionally implemented in this type of planning, associating each forest unit with one or more binary and/or integer variables. However, the higher number of binary/integer variables used in a model, the higher complexity and processing time to find an optimal solution. Given the need for improvements in computational performance, heuristic methods may present computational advantages to determine a feasible solution. In this context, the objective of this study was to solve a harvesting plan that aggregates adjacent stands to create blocks through the application of MILP and three different strategies of the Relax-and-Fix (R&F) heuristic. For that purpose, we used a hypothetical 1153-ha forest, subdivided into 100 stands of Pinus taeda L., as a database. The application of R&F, in general, proved to be efficient, especially when the R&F Backward solution was applied as an initial solution of the exact model. This scenario produced the best result of the study: a 28% decrease in computational time and the same spatial sequencing of stands as the MILP model.