Frequency-Tunable CNN Model for Radio Wave Propagation in Tunnel Environments

Machine learning (ML) techniques offer a promising tool for addressing the conventional trade-off between accuracy and efficiency in modeling wave propagation. Recent advancements have introduced an efficient ML model that can produce high-frequency field distributions from lower-frequency inputs, effectively mitigating the significant computational demands of high-frequency parabolic equation (PE) methods. Despite its efficacy, this model requires multiple trained models to predict each frequency combination, limiting its practical application. To address this, we propose an advanced tunable frequency conversion model that can predict high-frequency field distributions across continuously varying frequencies. Our numerical evaluations confirm the model's ability to accurately simulate field distributions from 1GHz to 3GHz in tunnels, showcasing its potential as a versatile tool for wave propagation prediction in such environments.