Meta-Learning for Boosting the Sensing Quality and Utility of FSO-Based Multichannel FBG Sensor System

In this article, we propose a novel meta-learning approach for improving the sensing quality of free space optics (FSO)-based multichannel fiber Bragg grating (FBG) sensor systems. Since it is difficult and time-consuming to collect large amounts of training data using real FBG sensor experiments, the proposed meta-learning can provide opportunities for reducing the amount of training data needed by learning from previous experience or knowledge. This makes our proposed meta-learning method suitable for scenarios, where a large amount of FBG sensor training data collection is challenging, limited, or expensive. Thus, we use a meta-learning approach to solve crosstalk problems and accurately predict the peak wavelength of each FBG sensor (<inline-formula> <tex-math notation="LaTeX">${S}_{n}$ </tex-math></inline-formula>) from the overlap spectra. The demonstrated results proved that using only a small amount of target data, the proposed meta-learning method can adapt more quickly and improve the central wavelength detection performance in the FBG sensor system than the traditional methods. Besides, the proposed FSO method addresses costs and topological limitations for the installation of fiber cables. Therefore, our proposed system is flexible, faster, cost-effective, adaptable, reduces the amount of training data needed, and can improve the accuracy of wavelength detection for FBG sensor systems.