In most metaheuristic approaches, particles are handled based on their fitness values without considering the distribution of solutions within the search space. Although this approach is simple, it causes particles to prematurely converge into a limited region of the search space due to the loss of diversity within the population. To address this limitation, this study proposes a metaheuristic approach in which particles are assigned to different search behaviors based on their Euclidean distance to the best solution. At each iteration, the population is divided into three groups: an exploitation set composed of the closest particles, an exploration set composed of the farthest particles, and a reference set composed of intermediate-distance particles. Two dedicated operators manage these groups: exploitation particles perform fine-grained refinement around the current best, whereas exploration particles search for new regions guided by randomly selected reference particles. In addition, an elitist acceptance mechanism ensures that only improved positions are retained, thereby promoting monotonic progress. This distance-based framework provides a distributed population of particles, where each particle is driven by its relevance to the optimal solution in the search space. This ensures a good diversity of solutions and prevents premature convergence and redundant search efforts. Experimental results on benchmark functions show that the method outperforms several State-of-the-Art metaheuristic algorithms in both solution quality and convergence behavior.
Cuevas et al. (Sat,) studied this question.