Machine learning-Assisted Object Monitoring supported by UWOC for IoUT
The research introduces an advanced fish movement tracking system that addresses critical challenges in underwater monitoring. By developing a lightweight sensor node integrated with an Inertial Measurement Unit (IMU) and underwater optical wireless communication (UWOC) modem, the system enables real-time data transmission and movement analysis. The technological innovation centers on sophisticated relay performance enhancement through software-based combining techniques like Majority Logic Combining (MLC), Equal Gain Combining (EGC), and Selection Combining (SC). These methods are intelligently integrated into the microcontroller to optimize communication reliability and signal processing in challenging underwater environments. A key contribution is the adaptive communication algorithm (ACA), which strategically exploits combining techniques to improve underwater wireless optical communication performance. The system's prototypes underwent rigorous testing in a 4-meter water tank, validating its feasibility and practical applicability. The research further advances underwater monitoring by developing machine learning models—including LSTM, Spatial Attention, RNN, Transformer, and GRU—trained on comprehensive IMU sensor data. These models analyze complex fish movement parameters, predicting acceleration states with remarkable precision and offering unprecedented insights into aquatic behavior.
Dec 5, 2024
