Deep Learning-Based Water Level Sensing With Interference Suppression for ISAC Systems

Effective disaster management and urban resilience are critical components of a smart city. In particular, flood monitoring has seen significant advancements driven by emerging technologies. While traditional flood monitoring methods persist, innovative solutions leveraging mobile communication signals are augmenting and, in some cases, supplanting them. In this letter, we propose a water level sensing scheme using integrated sensing and communication (ISAC) signals. To suppress interferences, our scheme uses beam scanning methods to generate two types of signals: one is used to estimate the water level, and the other one is used for communications. Machine learning technologies are used throughout the estimations. By exploiting the non-symmetric property of multiple received signals’ amplitude, we propose a “speed predictor” to obtain the variation speed of the water level. We then propose a multi-layer perceptron to estimate the water levels. Both experiments and numerical simulations are conducted and verify our proposed water level sensing scheme.