Identification and Robust Control of a Flapping Wing System Using Fuzzy Wavelet Neural Networks

Abstract

This study explores the utilization of a fuzzy wavelet neural network for identifying the dynamic behavior of a time periodic flapping wing system. Additionally, we propose a combination of adaptive sliding mode learning algorithm and genetic algorithm to effectively adjust the network parameters. The identification process specifically focuses on the longitudinal mode and involves three inputs and three outputs. To generate the necessary input and output data for the network, we have developed a simulation platform using MATLAB software. The accuracy of the data has been validated through comparison with ADAMS simulations. Moreover, we have ensured the stability of the learning algorithm by validating it with the Lyapunov stability theorem. The key contributions of this study include the simulation of a hummingbird that incorporates the effects of wing mass, as well as the introduction of a hybrid approach that combines genetic algorithms and sliding mode for network training purposes. This innovative methodology significantly enhances the performance of the network, resulting in improved convergence rates when compared to previous wavelet networks. Based on the results, we conclude that the designed network exhibits superior performance and is particularly well-suited for real-time applications. The combination of fuzzy wavelet neural network, sliding mode, and genetic algorithm offers a robust and efficient solution for accurately identifying the dynamic behavior of the flapping wing system.