Improved Egret Swarm Optimization Algorithm with Multiple Strategies
Abstract
The Egrets Swarm Optimization Algorithm is a recently proposed heuristic algorithm that simulates the hunting behavior of egrets. To address the limitations of the original algorithm, such as insufficient development capability and decreased population diversity, a multi-strategy improved Egrets Swarm Optimization Algorithm is proposed. First, in the population initialization phase, Logistic chaotic mapping is introduced to generate chaotic sequences, enriching population diversity. Next, a dynamic perception factor is introduced to replace the step size factor in the idle strategy, which allows for more effective exploration and discovery of potential optimal solutions. Furthermore, to increase the breadth and depth of exploration, a crayfish foraging strategy is incorporated into the random walk strategy, and a roulette wheel strategy is added to the encircling mechanism to enhance the algorithm’s search ability and avoid ineffective actions. Additionally, a distribution estimation strategy and a new exploration and exploitation strategy based on whale spiral ascent are introduced to improve the overall efficiency and functionality of the algorithm. Finally, testing on 20 classical functions shows that the improved algorithm enhances optimization performance. The algorithm is also applied to solve engineering constraint problems, demonstrating its practicality.
Keywords
Download Options
Introduction
In recent years, with the increasing complexity of problems and the ambiguity of the final results, the demand for optimization algorithms has grown. Optimization problems can be represented as a continuous or combinatorial design search space, where the process involves finding the maximum or minimum of a function.
Metaheuristic algorithms are a class of heuristic methods based on natural phenomena or species behaviors [1], designed to solve optimization problems by simulating the behaviors of organisms or species in nature [2]. These algorithms typically exhibit some degree of randomness and adaptability, enabling them to search for optimal or near-optimal solutions within the search space [3]. Their inspiration comes from various biological phenomena, such as animal collective behavior, plant growth patterns, and microbial reproduction methods. According to the No Free Lunch theorem [4], many metaheuristic algorithms have emerged, including the Zunhai Qiao algorithm [5], Artificial Bee Colony algorithm [6], Butterfly Optimization Algorithm [7] , Grasshopper Optimization Algorithm [8], Golden Sine Algorithm [9], Slime Mold Optimization Algorithm [10], Seagull Optimization Algorithm [11], Sparrow Search Algorithm [12], and Teaching-Learning-Based Optimization Algorithm [13] . Optimization problems are widespread across various fields, including engineering optimization, economics, logistics planning, machine learning, and artificial intelligence. Traditional optimization methods often face difficulties in solving complex high-dimensional, nonlinear, multimodal problems, whereas metaheuristic algorithms can effectively address various complex issues in the real world.
Conclusion
The Egrets Swarm Optimization Algorithm (ESOA) is a new metaheuristic algorithm introduced in recent years. It features a straightforward and easy-to-understand principle, is user-friendly, and is suitable for integration with other metaheuristic algorithms to address complex problems involving high dimensions or multiple optima. To enhance the efficiency of this algorithm, this paper proposes an improved version of the Egrets Swarm Optimization Algorithm, known as the Improved Egrets Swarm Optimization Algorithm (IESOA).
This improvement incorporates a chaotic local search strategy to boost the performance of the original ESOA algorithm. To address the issue of single strategy blindness in parallel algorithms, it introduces the "lost phase" and "roulette wheel" strategies from the crayfish algorithm. Additionally, to increase population diversity, a new parallel strategy is added to the original ESOA algorithm, which includes distribution estimation algorithms and spiral ascent strategies. This new strategy helps better guide the population towards more optimal solutions.
Comparative experiments with 20 classic benchmark functions and nine other intelligent optimization algorithms demonstrate that the performance of the IESOA algorithm has significantly improved over the ESOA algorithm. It is hoped that the IESOA algorithm can be applied to more engineering problems in the future, providing a feasible solution approach for addressing various practical issues.