Preliminary flight conflict detection using speed obstacle method based on SVM

This paper proposes a flight conflict detection method combining the Velocity Obstacle (VO) approach and Support Vector Machine (SVM). In a scenario with 1000 aircraft within a 1000×1000 km range and three altitude layers, ellipsoid models are used as protective zones for the aircraft. Positions and velocities of the aircraft are randomly generated, and potential conflict pairs are filtered using the VO method to train the SVM. The Radial Basis Function (RBF) kernel is employed in the SVM to identify high-probability conflict pairs, which are then confirmed by the VO method. Experimental results demonstrate that this method significantly enhances the accuracy and timeliness of conflict detection, effectively controlling both the false alarm rate and error rate.