Weevil Optimization Algorithm

To address the common problems of multimodality, strong nonlinearity, and local extremum traps in complex high-dimensional continuous optimization problems, this paper proposes a swarm intelligence optimization method based on the behavioral mechanism of weevils. The algorithm uses the long-snout probing behavior of weevils as a biological prototype, introducing a curvature-sensing detection mechanism, a nonlinear energy-driven convergence system, an information entropy adaptive adjustment mechanism, a dynamic topological neighborhood learning structure, and a fractal diffusion reproduction mechanism, forming a unified optimization framework coupling geometric structure, information statistics characteristics, and nonlinear energy dynamics. By constructing a diagonal curvature approximation tensor, local geometrically sensitive anisotropic step size adjustment is achieved; a nonlinear convergence dynamic model is constructed using individual energy variables; the population distribution state is characterized and the exploration intensity is adjusted using swarm information entropy; the multimodal adaptability is improved through dynamic neighborhood topology; and the global escape capability is enhanced through fractal perturbation. Under continuous-time limits, the algorithm can be represented as a nonlinear, anisotropic, and energy-coupled stochastic dynamic system. Theoretical analysis shows that the method possesses asymptotic stability under certain conditions. This paper constructs a complete mathematical model and systematically analyzes the dynamic structure of the algorithm.