Air Traffic Trajectories Segmentation Based on Time-Series Sensor Data

Abstract

ATC is a critical area related with safety, requiring strict validation in real conditions (Kennedy & Gardner, 1998), being this a domain where the amount of data has gone under an exponential growth due to the increase in the number of passengers and flights. This has led to the need of automation processes in order to help the work of human operators (Wickens et al., 1998). These automation procedures can be basically divided into two different basic processes: the required online tracking of the aircraft (along with the decisions required according to this information) and the offline validation of that tracking process (which is usually separated into two sub-processes, segmentation (Guerrero & Garcia, 2008), covering the division of the initial data into a series of different segments, and reconstruction (Perez et al., 2006, Garcia et al., 2007), which covers the approximation with different models of the segments the trajectory was divided into). The reconstructed trajectories are used for the analysis and evaluation processes over the online tracking results. This validation assessment of ATC centers is done with recorded datasets (usually named opportunity traffic), used to reconstruct the necessary reference information. The reconstruction process transforms multi-sensor plots to a common coordinates frame and organizes data in trajectories of an individual aircraft. Then, for each trajectory, segments of different modes of flight (MOF) must be identified, each one corresponding to time intervals in which the aircraft is flying in a different type of motion. These segments are a valuable description of real data, providing information to analyze the behavior of target objects (where uniform motion flight and maneuvers are performed, magnitudes, durations, etc). The performance assessment of ATC multisensor/multitarget trackers require this reconstruction analysis based on available air data, in a domain usually named opportunity trajectory reconstruction (OTR), (Garcia et al., 2009). OTR consists in a batch process where all the available real data from all available sensors is used in order to obtain smoothed trajectories for all the individual aircrafts in the interest area. It requires accurate original-to-reconstructed trajectory’s measurements association, bias estimation and correction to align all sensor measures, and also adaptive multisensor smoothing to obtain the final interpolated trajectory. It should be pointed out that it is an off-line batch processing potentially quite different to the usual real time data fusion systems used for ATC, due to the differences in the data processing order and its specific 2