Data-driven control framework using fractional order singular optimal control and optimized metaheuristic algorithms

Abstract

As the demand for advanced healthcare systems increases with the aging population, this paper introduces a novel data-driven control framework for constrained systems. The framework integrates signal processing algorithms with the optimal control of fractional order singular systems. Data collection was performed using a master-slave structure, while the classification process included preprocessing, window selection, feature extraction, and feature selection conducted through a genetic algorithm. We used machine learning algorithms, fuzzy wavelet neural networks using optimized metaheuristic algorithm, and convolutional neural network-long short-term memory (CNN-LSTM) for classification. We first decomposed both time-invariant and time-varying systems for the controller design to simplify the control process. This was followed by eliminating infinite modes, allowing for more efficient system control. We developed a novel linear method based on orthogonal functions to address the presence of both left and right fractional-order derivatives. The proposed framework's practicality was validated through its application in a rehabilitation system. Results indicated that electromyography (EMG) signals effectively classified movement states when combined with machine learning algorithms. In contrast, electroencephalogram (EEG) signals were better suited for classifying mental states. For movement classification using EEG signals, the fuzzy wavelet neural network and optimized CNN-LSTM emerged as the most effective methods. Among the orthogonal functions, the Chebyshev polynomial delivered the best performance, further confirming the robustness of our approach.