An Hypervolume based constraint handling technique for multi-objective optimization problems

Formulation of structural optimization problems usually leads to the individuation of one or more objective functions to be minimized under different constraints. Many multi-objective evolutionary algorithms are approached by a Pareto-compliant ranking method, where no a priori information on the problem is needed and the concept of non-dominated solutions is used. In this paper a constraint handling technique based on the concept of hypervolume indicator is presented. Initially proposed to compare different multi-objective algorithms hypervolume indicator is the only single set quality measure to reflects the dominance of solution’s sets. The constraint handling technique proposed use an extension of stochastic ranking approach for single-objective optimization problem to multi-objective ones. The extension proposed use the hypervolume indicator to compares different solutions and is tested on a structural constrained multi-objective problems. Results show the suitability of the proposed approach.