A 6G Integrated Sensing and Communication Framework for Railway Intrusion Detection and Collision Prediction

As 6G networks evolve, they are expected to integrate sensing capabilities directly into communication infrastructure—a paradigm known as Integrated Sensing and Communication (ISAC). This paper explores whether the Channel State Information (CSI) inherent in 6G wireless signals can be used to detect intruders on railway tracks and estimate their kinematics (position, velocity, and time-to-collision) to prevent accidents.

The authors created a high-fidelity 3D-rendered railway environment using Blender and simulated wireless propagation using the Sionna radio simulator. They generated 22,695 CSI matrices representing various intruder movements within a defined 'danger zone' (3 meters from the track). The raw CSI data underwent a preprocessing pipeline involving subcarrier selection, static background removal, and temporal denoising. This processed data was then fed into a hybrid machine learning model, which combined a 3D Convolutional Neural Network (3D CNN) for spatial feature extraction and a Bidirectional Long Short-Term Memory (BiLSTM) network to capture temporal dynamics.

The proposed model demonstrated high effectiveness on the synthetic dataset. It achieved a 99.57% classification accuracy for detecting the presence of an intruder in the danger zone. Furthermore, the model successfully performed regression tasks to estimate the intruder's kinematic parameters, achieving a combined Mean Absolute Error (MAE) of 0.4240. These results suggest that standard 6G communication waveforms can be repurposed for reliable, real-time safety monitoring without requiring dedicated radar hardware.

Railway safety is often compromised by the difficulty of monitoring vast, remote stretches of track. Current solutions like cameras, LiDAR, or infrared sensors are often limited by line-of-sight requirements, environmental conditions, or high costs. By leveraging existing 6G communication infrastructure for sensing, this framework offers a potentially scalable and cost-effective solution for continuous, all-weather intrusion detection.