Machine Learning Enhanced Multi-Sensor Fusion for Air Traffic Surveillance
Zoé Besson, Mengwei Sun, Ivan V. Petrunin, Andrew Hill, David E. Johnson · 2025
The rapid proliferation of Unmanned Aerial Vehicles (UAVs) in civilian airspace has introduced significant challenges to air traffic management, including collision risks and unauthorized airspace intrusions. Existing sensing technologies, such as radar, ADS-B, cameras, and acoustic sensors, face individual limitations related to environmental constraints, spoofing risks, and range restrictions, rendering them insufficient for robust UAV detection and tracking. This paper presents a novel machine learning-enabled multi-sensor fusion framework to address these challenges. The proposed approach integrates acoustic, visual, radar, and ADS-B sensor data through a three-tier fusion architecture. The first tier fuses acoustic and visual data for enhanced detection capabilities, the second tier incorporates radar data using Extended Kalman Filters (EKF) for trajectory estimation, and the third tier employs high-level fusion for temporal data fusion with ADS-B. Extensive simulations and performance evaluations demonstrate that the proposed method significantly improves detection accuracy and tracking robustness compared to traditional single-sensor approaches. This work represents the next step toward enabling safer and more reliable UAV operations in complex, shared airspace environments.