Covariance matrix adaptation driven dynamic multi-population colony predation optimizer: Insights, qualitative analysis, and constrained engineering optimization

Abstract

The Colony Predation Algorithm (CPA) is a straightforward population-based algorithm with few control parameters. Nevertheless, its initial design has limitations, including a tendency for local optimization and limited search ability, leading to subpar solutions. In response to these concerns, a novel approach named the Covariance Matrix Adaptive-Driven Dynamic Multi-Population Colony Predation Algorithm (ICPA) is introduced. Qualitative analysis experiments are carried out to determine ICPA's feasibility, including historical search trajectory analyses and balanced diversity assessments. Additionally, a comparative study involving 11 well-known algorithms, 11 state-of-the-art algorithms, and four champion algorithms (EBOwithCMAR, SPS_L_SHADE_EIG, LSHADE_cnEpSi, and LSHADE) using the IEEE CEC 2014 test suite confirmed its superior optimization capabilities. Statistical tests consistently rank ICPA first on the Friedman test (with scores of 2.17, 3.37, and 2.27) and demonstrate its outperformance of state-of-the-art algorithms in at least 40 % of tested functions in the Wilcoxon sign-rank test. The Bonferroni Dunn post-hoc statistical test reveals that ICPA significantly outperforms 61.53 % of the compared algorithms. Additionally, the efficacy of ICPA was evaluated on seven constrained real-world engineering problems, encompassing gear train design, speed reducer design, multi-disc clutch and brake design, cantilever beam design, three-bar truss design, I-beam design, and combined economic emission dispatch. The experimental outcomes underscore the potential of ICPA in addressing practical engineering challenges, thereby validating its optimization effectiveness. Utilizing extensive experimentation and comparative analyses, the feasibility, superiority, and effectiveness of ICPA have been substantiated, encompassing both qualitative and quantitative perspectives.