Deep Learning-Based Jammer Localization in Railway Wireless Networks Using Multi-AOA Fusion and Kalman Filtering

Railway wireless communications play a critical role in transmitting train control and dispatch commands, where high reliability and low latency are essential for operational safety. However, characterized by high mobility and increasing electrification, railway systems are susceptible to electromagnetic interference (EMI), potentially leading to service disruptions and safety compromises. Intentional EMI (IEMI), or jamming, which is deliberately and maliciously generated, further exacerbates the vulnerability of these critical communication networks, thereby making jammer detection and localization increasingly important. This paper proposes composite models reflecting realistic railway scenarios, dividing the localization approach into two steps: Angle of Arrival (AOA) identification and 3D geometric triangulation. A deep learning algorithm integrating convolutional neural networks (CNN) and long short-term memory (LSTM) networks is developed to estimate AOA by learning spatial, temporal, and frequency characteristics of received signals. Subsequently, multiple AOAs are fused, and a Kalman filter-based 3D geometric triangulation method is employed to determine the jammer location. The proposed method achieves real-time AOA estimation with an inference latency of just 2.69 ms and an angular error below 0.71°, thereby meeting the stringent latency and reliability requirements of railway applications. Multi-AOA fusion with Kalman filtering achieves a localization accuracy within a 2-meter radius for 92% of the data, reducing errors by 27% compared to state-of-the-art methods. Simulation results demonstrate the adaptability of the proposed method across four representative railway scenarios at train speeds up to 500 km/h.This work has been submitted to the IEEE Journal on Selected Areas in Communications for possible publication. Copyright may be transferred without notice, after which this version may no longer be accessible