A Multi-objective Evolutionary Hybrid Optimizer

A new hybrid multi-objective, multivariable optimizer utilizing Strength Pareto Evolutionary Algorithm (SPEA), Non-dominated Sorting Differential Evolution (NSDE), and Multi-Objective Particle Swarm (MOPSO) has been created and tested. The optimizer features automatic switching among these algorithms to expedite the convergence of the optimal Pareto front in the objective function(s) space. The ultimate goal of using such a hybrid optimizer is to lower the total number of objective function evaluations in multiobjective multi-extrema optimisation problems. The SPEA is an elitist evolutionary algorithm proposed by Zitzler and Thiele. The algorithm introduces elitism by maintaining an external population for the non-dominated set at each of the iterations. This algorithm uses the elite members in genetic operation to steer the population towards optimal region in multi-dimensional search space. It has the in-built clustering technique, which helps in creating a better spread of the non-dominated solutions. It also supports multi-variables and multi-objective functions. The algorithm has been tested for standard multi-objective test functions and given acceptable results. The MOPSO algorithm is a modified version on the algorithm proposed by Eberhart and Kennedy. Although this implementation of the original algorithm is primitive, MOPSO supports multi-variables and multi-objective functions. The algorithm has shown the ability to find acceptable optimal Pareto fronts for numerous standard multi-objective test functions. George S. Dulikravich, Ramon J. Moral and Debasis Sahoo.