A Measure of Diversity for Metaheuristic Algorithms Employing Population-Based Approaches

In Metaheuristic Algorithms (MA), balancing exploration and exploitation is a widely recognized challenge and an ongoing research issue within the MA community. Apart from the specific parameters inherent to these algorithms, an alternative method to regulate the Exploration–Exploitation Balance in MA involves employing a Diversity Metric (DM) as a guiding mechanism. However, this approach faces two main challenges: the computational expenses associated with its implementation and its efficacy in accurately representing the true diversity within the population. This chapter introduces a novel DM tailored for real-coded candidate solutions. The proposed approach utilizes surrogate hypervolumes to assess the spatial distribution of individuals within the population. In a comparative analysis with five diversity metrics discussed in existing literature, our proposition demonstrates comparable stability, sensitivity, and robustness outcomes, particularly in the presence of outliers, all achieved without significant increases in computational overhead.