Electromagnetic spectrum sensing experiments and prediction models in the marine environment

There is growing demand for accurate ocean environmental monitoring and spectrum sensing technology, and traditional single-sensor systems are no longer enough to meet the technical needs of multidimensional parameter collection and analysis. To address this challenge, we developed a custom maritime monitoring system designed to achieve efficient, stable, and continuous observations of both oceanic meteorological conditions and electromagnetic spectrum distribution.

This intelligent platform with multimodal sensing capabilities integrates five main functional modules: a positioning and attitude sensing unit, a meteorological parameter collection unit, a data storage and management unit, a remote communication module, and an electromagnetic spectrum reception module. It allows for the synchronized collection and local processing of more than ten key parameters, including wind speed, wind direction, attitude angles, power supply data, and electromagnetic signal strength. Structurally, the platform features a modular, low-power design, making it suitable for long-distance deployment and mid- to long-term operation, greatly enhancing deployment flexibility and data continuity on the ocean surface. The principles and procedures of electromagnetic signal reception are described in detail here. Regarding the field measurement process, sensor attitude was examined, and a preliminary analysis was performed on how platform orientation and meteorological factors (such as wind speed and direction) influence signal strength. A compensation model for platform attitude was then proposed. To evaluate the platform's performance, it was deployed in a typical dynamic maritime environment to continuously record the received signal strength indicator (RSSI) within a target frequency band. Real-time attitude data (pitch, roll, and yaw) and meteorological variables (wind speed and direction) were recorded simultaneously. Using the collected data, both linear and nonlinear multivariate regression models were developed to analyze how platform motion and environmental disturbances impact signal reception. Linear models proved insufficient to capture the complexity of attitude and meteorological influences on signal strength, leading to the adoption of a random forest regression algorithm, which demonstrated strong performance in handling nonlinear and multivariable interactions.

Results reveal a significant negative correlation between yaw angle and RSSI, with wind direction deviation also exerting a moderate influence on signal strength. When compared to the XGBoost and SVR algorithms, the random forest model showed superior accuracy and stability, achieving an R² value greater than 0.84, indicating a high explanatory power of platform attitude and environmental factors on signal strength variations.

This study is the first to incorporate multi-source heterogeneous environmental sensing data into signal strength modeling. It introduces a novel analytical framework tailored for dynamic marine environments that improves spectrum sensing accuracy and stability on mobile platforms. Both theoretical insights and experimental validation are provided. The proposed system and methodology demonstrate strong scalability and broad applicability, suitable for uses such as marine meteorological monitoring, maritime electromagnetic interference detection, and wireless communication link assessment. Additionally, this system will serve as a valuable educational tool in universities, supporting both teaching and practical training.