Currently, the acquisition and application of multimodal sensing information are deeply integrated with artificial intelligence, becoming a leading hotspot in the field of intelligent perception. With the help of AI, especially deep learning methods, massive and heterogeneous sensor data (such as LiDAR point clouds, spectral images, radar signals) are no longer passive collected physical quantities 12345, but have become the core assets driving intelligent decision-making. AI endows multimodal fusion with powerful analytical capabilities, enabling the extraction of high-dimensional features from complex environments and achieving more accurate target recognition, situation awareness, and prediction 6789. In fields such as smart cities, autonomous driving, and industrial Internet of Things, this combination not only breaks through the perception limitations of a single sensor, but also promotes the system's transition from perception to cognition, providing key technical support for achieving autonomous intelligence and adaptive control. ) presents a novel method for optical fiber plane curve reconstruction using cubic spline interpolation and tangent angle recursion. Strain data from sensors yields discrete curvature values, which are smoothed via interpolation. Coordinates are then calculated recursively for precise reconstruction. With 50 sampling points, mean absolute error reaches 0.000892 m, significantly outperforming lower sampling rates and validating the method's feasibility. Liu et al. (https://doi.org/10.3389/fphy.2025.1638385) presents a broadband metasurface antenna designed using characteristic mode analysis. By loading parasitic patches and etching slots based on mode optimization, the antenna achieves improved impedance matching and enhanced high-frequency gain. The optimized higher-order modes exhibit broadside radiation, significantly boosting realized gain. This CMA-based approach provides an effective strategy for designing high-performance, flat-gain antennas.This Research Topic brings together cutting-edge research on the acquisition and application of multimodal sensing information, fully reflecting the development trend of deep integration with artificial intelligence. In the field of gas sensing, researchers have developed a photoacoustic spectroscopy-based sensor for dissolved acetylene detection in transformer oil, achieving a high sensitivity of 50 ppb. Systematic reviews are provided on the progress of quartz-enhanced photoacoustic spectroscopy (QEPAS) and gas monitoring technologies for lithium-ion battery thermal runaway, with special emphasis on the electromagnetic interference immunity of optical fiber sensors. In the field of LiDAR, an innovative point cloud down-sampling method based on fuzzy C-means clustering was proposed to effectively preserve geometric features. An airborne LiDAR system suitable for high-speed flight platforms up to 120 km/h was developed, enabling real-time power line recognition and terrain assessment. Furthermore, research was conducted on accuracy optimization for optical fiber curve reconstruction and the design of a broadband metasurface antenna based on characteristic mode analysis. These achievements demonstrate a transition from single sensors to multimodal fusion, and from data acquisition to intelligent cognition, providing key technical support for applications such as autonomous driving, smart cities, and the industrial Internet of Things.
Yin et al. (Tue,) studied this question.