4D Trajectory Prediction Based On Data Mining Methodological Study

One of the core operating mechanisms in China's New Generation of Air Traffic Management systems(CNGATM)is based on track-based operation(TBO).Track-based operation requires coordinated decision-making track management in high-density,high-complexity airspace.Track prediction is used to assist the air traffic control system to predict aircraft movements in advance,and predict the flight time and arrival key of the flight through 4D(four-dimensional)tracks.At the moment of the node,macro control of the entire air traffic flow can also detect and deal with potential flight conflicts in time.With the advent of the data age,how to extract and analyze the organization pattern and change law of traffic flow from the monitoring data of multi-source heterogeneous and massive air traffic control is also a problem that needs to be solved urgently in 4D track prediction technology.Aiming at the 4D track prediction problem,a 4D track prediction method based on data mining and CURE clustering algorithm is constructed.Firstly,based on the ADS-B historical radar data,a large number of data samples are obtained through data preprocessing.Aiming at the problem that the traditional prediction algorithm can only process a small amount of data,this paper uses the ACCESS database to process the data with redundancy and difference,and obtains a smooth and complete normalized trajectory set by the similarity calculation and filtering of the Euclidean distance.For most current track predictions,only clusters with three-dimensional coordinates are considered,and the influence of flight parameters such as heading change and vertical height on flight clustering results is not considered.In order to avoid this problem,CURE algorithm cluster analysis is used.Make up for the neglect of height and heading in previous studies.In the clustering process,it is found that the terminal area track clusters of the same flight have different classifications,so the feature track is extracted in a segmented form for the flight path prediction of the flight.For the flight conflict in the terminal area,the 4D pane flight conflict screening algorithm is used for collision detection,and it is optimized by the geometric method of flight conflict precision prediction algorithm.The case results show that the optimized algorithm reduces the false flight conflict alarm by 17.5%.For the flight change of the local emergency situation,this paper makes local prediction from the latitude and longitude offset,time deviation and horizontal speed deviation,which is suitable for the prediction of the emergency navigation track.The example uses the ADS-B track of the CDG4651 flight from January to May 78.The prediction results are compared with the actual flight trajectory of a certain day.The results show that the prediction error is in a small range.