Analysis and evaluation of the R-ratio heuristic for forest scheduling

Abstract

The R-ratio heuristic for forest scheduling (Rodriguez, 1994) solves large scale 0-1 integer programming versions of the Type I harvest scheduling problem (Johnson and Scheurman, 1977). This paper reviews its foundations and describes its evolution until now. Heuristic scheduling system patterns are used to represent the problem, to describe the evaluation function and to monitor the solution process. Its logic and elements are presented based on some artificial intelligence (AI) principles proposed to build heuristic processes. An AI approach based on intelligent agents provides the basis to analyze the R-ratio's (i) escape strategy from local optima and (ii) its hybrid A*-greedy strategy to the solution search. AI concepts are also utilized to evaluate performance indicators of efficacy, measured by the proximity to the optimal solution of the non-integer linear programming relaxed version of the same problem, and efficiency, measured by penetrance and space complexity. For the test problems, the R-ratio strategy to escape from local optima proved efficacious given that several feasible solutions with objective function values bellow the range of 0.5% were obtained. And the R-ratio approach to find feasible solutions also proved efficient given its focus on a low cost strategy to select path searches.