Asynchronous Track to Track Association Using Dynamic Time Warping

Abstract

Target tracking is one of the basic feature of modern day radars. In distributed multi sensor multi-target environment, the data of local radar trackers is sent to a centralized location (fusion center) for processing and displaying complete sensors picture. This data is usually asynchronous due to different sampling rates of radars and communication delays. At fusion center (central node) the tracks from various radars are compared based on distance metrics and thresholds for subsequent Track-to-Track Association (T2TA) and fusion process. Currently employed methodologies for track-to-track asso ciation are usually based on time alignment of tracks before association process using recursive estimation and state covariance data, however as a result of synchronization and estimation process the errors in data spreads. This research proposes application of Dynamic Time Warping (DTW) algorithm for track to track association problem.The tracks data does not require time alignment and can be compared directly. The use of DTW for association of aircraft tracks is relatively a new approach, and thus can be an addition to existing methodologies. During the study different simulated tracks by multiple radars were analyzed and Dynamic Time Warping (DTW) algorithm was implemented for correct T2TA. A sliding window technique has been proposed to speed up the DTW calculations and to offer real time tracks association. The tracks have been plotted in 3D Cartesian coordinates frame and corresponding DTW cost has been calculated using dynamic programming matrix. The DTW cost is compared with a threshold for finding the correct association. A complete system model has been defined to incorporate measurement errors in slant range (r), azimuth (θ) and elevation (φ) angles. Simulation has been performed in Matlab and the results show effective asynchronous track-to-track association using DTW.