Mandrill Optimization Algorithm

Abstract

This paper introduces the Mandrill Optimization Algorithm (MOA), a novel multi-swarm and derivative-free metaheuristic algorithm inspired by the individual intelligence and dominance hierarchy observed in mandrill hordes while foraging. MOA aims to address the challenges of slow convergence speed and trapping in local optima, particularly in high-dimensional optimization problems. The algorithm utilizes multiple innovative mathematical models to accurately simulate the movement and foraging behaviors of mandrills during the day. MOA distinctly separates the exploration and exploitation phases, adjusting their balance smoothly based on the ratio of elapsed time of the day. MOA’s performance is rigorously evaluated in two stages. In the first phase, it is benchmarked against 40 mathematical test functions, including fixed-dimensional, multimodal, composite, and high-dimensional unimodal functions, to assess convergence speed, avoidance of local minima, and exploration-exploitation trade-offs. In the second phase, our proposal is applied to five real-world optimization problems to validate its practical utility. Comparative analysis of recent and traditional metaheuristic algorithms highlights MOA’s robust performance. While the results on high-dimensional benchmarks are competitive, this proposal demonstrates significant superiority on multimodal benchmarks. These findings underscore its potential as an effective and efficient optimization tool for complex problem domains.

Keywords: Optimization Algorithm, Mandrills, Mandrill Optimization Algorithm, Metaheuristic, Mathematical Model, Heuristic Algorithm