Blended Rank Evolutionary Algorithm for the Constrained Multiobjective Crop Rotation Problem

In a constrained multiobjective problem, solutions can be mapped onto three spaces: decision variable space, objective space, and constraint space. Blended Rank Evolutionary Algorithm uses measures from all three spaces and dynamically blends them together into a final fitness score for use in an evolutionary algorithm. Results on the highly constrained, multiobjective "nonlinear crop rotation" problem show that BREA reliably finds better quality non-dominated fronts than the popular algorithm NSGA-II. The difficulty of the nonlinear crop rotation problem leaves room for improvement in both algorithms.