The Maximin Fitness Function for Multi-objective Evolutionary Computation: Application to City Planning

摘要

A new fitness function, known as the maximin fitness function, is presented for multi-objective genetic algorithms. This fitness function directs genetic algorithms towards final generations that are both close to the universal Pareto front and diverse. The performance of a genetic algorithm with the maximin fitness function as well as with the traditional Pareto-ranking fitness function is compared on a real-world test problem from city planning.