A Probabilistic Conflict Detection Algorithm Based on Ensemble Learning in Free Flight

Abstract

The continuous increase in air traffic flow has put forward higher requirements on the conflict detection accuracy and the ability to handle with a large number of aircraft. The conflict detection algorithm based on ensemble learning is put forward. Firstly, aircraft conflict model is established for the purpose of selecting flight data sets. Secondly, the current positions, speed vectors, look-ahead time, estimated turning time and turning angles are extracted as characteristic quantities which are inputted to train the basic classifiers, and a meta-data sets are obtained. Afterwards, support vector machine is used as the second-level classifier, the meta-data sets are regarded as new characteristic quantities which is used to train the stacking meta classifier. Lastly, the conflict probability is solved by the Sigmoid function mapping method. Simulation results show that this algorithm has a high accuracy for conflict detection, improve the problem of high false alarm rate, and is suitable for turning flight.

Keywords

• conflict detection
• ensemble learning
• stacking meta learning
• air traffic management