Great Wall Construction Algorithm: A novel meta-heuristic algorithm for engineer problems

Abstract

In recent years, the optimization community has witnessed a surge in the popularity of population-based optimization methods. However, many of these methods suffer from various shortcomings, including unclear performance characteristics, incomplete validation, excessive reliance on metaphors, inadequate exploration and exploitation components, and compromised trade-offs between exploration and exploitation in real-world scenarios. As a result, users often find themselves needing to extensively modify and fine-tune these methods to achieve faster convergence, stable balance, and high-quality results. To shift the optimization community’s focus towards performance rather than metaphorical changes, we propose a general population-based optimization technique called the Great Wall Construction Algorithm (GWCA). This study presents GWCA as a simple yet robust method with competitive performance for efficiently solving constrained and unconstrained problems. GWCA draws inspiration from the competition and elimination mechanisms observed among workers during the construction of the ancient Great Wall. It introduces a mathematical model of the labor movement to simulate the algorithm’s dynamics. Unlike other methods that employ multiple models to generate new solutions, GWCA randomly assigns a single predefined motion model to each worker in every iteration. This unique approach showcases GWCA’s dynamic nature, simple structure, high convergence performance, and ability to deliver satisfactory solution quality, thus outperforming existing optimization methods in terms of efficiency. To validate GWCA, we conduct extensive comparisons with popular and advanced algorithms on the IEEE CEC 2017 benchmark suite across different dimensions (D = 10, 30, 50, 100). Additionally, GWCA is applied to solve 16 constrained engineering problems and 6 NP-Hard problems, demonstrating its applicability in handling constrained and complex nonlinear problems. Finally, we compare GWCA’s optimized solutions with those obtained from 33 advanced meta-heuristic algorithms, including the winner of CEC 2017. The results confirm the effectiveness of the proposed optimizer in solving a wide range of single-objective problems, surpassing popular base optimizers, advanced variants of existing methods, and several CEC winners. We present GWCA as an open-source population-based method that can serve as a standard optimization tool across various domains of artificial intelligence and machine learning. It exhibits a range of exploratory and exploitative features, offering high performance and optimization capabilities. The method is highly flexible, scalable, and can be further extended in terms of structure and application to accommodate diverse forms of optimization scenarios. https://github.com/guangian/Great-Wall-Construction-Algorithm-a-novel-meta-heuristic-algorithm-for-global-optimization.