On a new constraint handling technique for Multi-Objective Genetic Algorithms
Jin Wu, Shapour Azarm · 2001
A new constraint handling technique is developed to work with Multi-Objective Genetic Algorithms (MOGAs). This technique is based on a primary-secondary fitness assignment scheme, one that uses both individuals' fitness and matching. A Pareto ranking scheme is used for the primary fitness assignment wherein no subjective and problem dependent parameters are used. Rules that take the concept of matching into account are used for the secondary fitness assignment. Some new set quality metrics are introduced and used for a comparison of the new technique with a previous approach. Due to the stochastic nature of MOGA, confidence intervals with a 95% confidence level are obtained for the quality metrics based on the randomness in the initial population. An engineering example, namely the design of a vibrating platform, is used for the comparison and demonstration of the new technique.