Use of the Multiobjective Technique NPGA to Handle Constraints in Genetic Algorithms.
Maestría en Inteligencia Artificial
University of Veracruz
Defended on August 24th, 2001. Honorific Mention
Adviser: Dr. Carlos A. Coello Coello
CINVESTAV-IPN, Depto. de Ing. Eléctrica, Sección deComputación
In this work, we propose an approach to handle constraints in evolutionary algorithms based on a evolutionary multiobjective optimization technique called Niched Pareto Genetic Algorithm (NPGA). NPGA uses tournament selection based on nondominance checking. A state-of-the-art of constraint handling in evolutionary algorithms is presented. Also, we describe the modifications performed to the NPGA to solve global optimization problems with constraints. After that, the results obtained are presented and compared against other approaches. Finally, we analyze the sensitivity of the approach to its parameters.
Document (.ps in spanish)
Document (zipped .pdf in spanish)