Adaptive Crow Search Algorithm Based on Population Diversity

Abstract

As an emerging metaheuristic optimization algorithm, the Crow Search Algorithm (CSA) has received widespread academic attention for its intuitive model and concise parameter settings. However, its inherent limitations, including the tendency to fall into local optima and insufficient convergence accuracy in solving complex optimization problems, have severely restricted its popularization and application in practical engineering scenarios. Aiming at the above shortcomings of the basic CSA, this thesis conducts in-depth and systematic research, proposes two targeted improved algorithms, and systematically verifies their theoretical performance and engineering application value.To address the performance bottleneck of the basic CSA, this thesis proposes an Adaptive Crow Search Algorithm (ACSA) based on population diversity. By constructing an adaptive framework with population diversity as the feedback signal, the algorithm achieves three core improvements. First, a dynamic dual-mode guidance mechanism is designed to intelligently switch between global exploration and local exploitation strategies according to the population diversity index. Second, the Sigmoid function is introduced to realize adaptive adjustment of flight length, enabling the parameter settings to match the real-time search state of the algorithm. Third, the golden sine strategy is adopted to optimize the individual position update rule, which is combined with a reflection boundary handling mechanism to enhance the iterative stability of the algorithm.