Model Identification and Attitude Control of Bionic Butterfly

The potential applications of bionic butterflies in both military and civilian domains are extensive, primarily due to their remarkable flexibility and concealment capabilities. This study addresses the challenges of modeling and attitude stabilization for these complex systems. The key contributions include a novel hybrid Radial Basis Function (RBF) - Multi-Layer Perceptron (MLP) neural network for efficient system identification, specifically designed for embedded deployment feasibility, and an enhanced fuzzy adaptive Active Disturbance Rejection Controller (ADRC) strategy enabling robust, decoupled control of pitch and heading angles. The RBF-MLP network effectively captures the approximate relationship between the wing up-down swing angles and the bionic butterfly's attitude angles. Building upon this identified model, the improved ADRC addresses the inherent time-varying characteristics of butterfly motion. Fuzzy logic integration within the extended state observer and nonlinear feedback facilitates adaptive parameter adjustment and enhances system robustness, achieving decoupled control over pitch and heading. Experimental results demonstrate the designed control system's rapid convergence for both pitch and heading angles, initially validating the efficacy of our proposed system identification model and control strategy for managing attitude within this innovative platform.