Beyond Bio-Inspired Algorithms: Using Bird Landing Dynamics to Design Adaptive Parameter Control in Metaheuristics

Abstract

Bio-inspiration has mainly been used to represent organisms, behaviors, or natural processes in the design of metaheuristics. This study proposes a different use: drawing on a biological phenomenon to model the temporal evolution of an internal parameter in an existing algorithm. Logarithmic Mean Optimization (LMO) is adopted as a case study, focusing on β, which scales the stochastic perturbation term and regulates the balance between exploration and exploitation. Inspired by the progressive transition observed during bird landing, a normalized arctangent trajectory controlled by m and k is proposed. Both hyperparameters were tuned through Bayesian optimization using the Tree-structured Parzen Estimator (TPE) implemented in Optuna. The 23 benchmark functions were divided into 12 tuning functions and 11 independent test functions. Nine β configurations were evaluated through 31 runs per function. The Friedman test showed significant differences among the variants (p=3.8269×10−10), and the proposed formulation achieved the best average rank (1.7273). Post-hoc Wilcoxon tests with Holm correction found significant differences against two of the eight alternatives. Overall, the results suggest that bio-inspiration, when used as a criterion for designing adaptive parameter-control mechanisms, can yield improvements and be considered a potential alternative in the design of new metaheuristic algorithms.