Multi-objective Genetic Algorithms

The goal of an optimization model is to identify the optimal values for decision variables with the objective of either maximizing or minimizing an objective function. When there is more than one objective function involved, finding the optimum solution(s) becomes a multi-objective optimization (MOO) task. Multi-objective optimization usually results in a set of optimal solutions called the Pareto optimal set. There are several algorithms available for finding the Pareto optimal set for MOO problems. Evolutionary algorithms (EAs), such as genetic algorithms, are stochastic search techniques that are designed to solve complex optimization problems. These algorithms work by processing a population of solutions in each iteration, making them particularly effective in MOO. Hence, multi-objective EAs have been very popular in the last three decades. Multi-objective genetic algorithms (MOGAs), in particular, have become the preferred heuristic method for solving MOO problems. This chapter first reviews multi-objective evolutionary and genetic algorithms and then presents the fundamental principles and design considerations of MOGAs such as encoding, crossover and mutation operators, fitness assignments, selection methods, and diversity preservation. Applications, future directions, challenges, and opportunities concerning MOGAs are also discussed.