2022 Volume 13 Issue 1 Pages 53-65
SAT constraints have been recently proposed as a flexible and effective method to specify the feasible space of binary multi-objective optimization problems and create scalable constrained problems. They derive from a variant of SAT problems, where the propositional logic formula composed of weighted 3-SAT clauses is split to create the desired number of equality and inequality constraints. This work studies the effectiveness of four constraint handling techniques solving 0-1 bi-objective knapsack problems extended with SAT constraints, namely the death penalty, constraints dominance, static penalty, and co-evolution. We aim to understand the scalability of the constraint handling techniques on problems with equality and inequality constraints, varying the feasibility ratio of the generated problems keeping constant the number of constraints and varying the number of constraints keeping constant the same feasibility ratio.
