A comparative analysis of trajectory similarity measures

Computing trajectory similarity is a fundamental operation in movement analytics, required search, clustering, and classification of trajectories, for example. Yet the range different but interrelated measures can be bewildering researchers practitioners alike. This paper describes systematic comparison methodical exploration measures. Specifically, this compares five most important commonly used measures: dynamic time warping (DTW), edit distance (EDR), longest common subsequence (LCSS), discrete Fréchet (DFD), (FD). The begins with thorough conceptual theoretical comparison. highlights similarities differences between connection six characteristics, including their handling relative versus absolute space, tolerance to outliers, computational efficiency. further reports on an empirical evaluation trajectories contrasting properties: data about constrained bus movements transportation network, unconstrained wading birds coastal environment. A set four experiments: a. creates measurement baseline by comparing single subjected various transformations; b. explores behavior network-constrained grouped based spatial temporal similarity; c. assesses respect known behavioral annotations (flight foraging oystercatchers); d. bird activity examine whether they are distinguishable solely patterns. results show that all instances both value ordering may sensitive choice measure. In general, were more able distinguish than differences. concludes high-level summary advice recommendations selecting using practice, conclusions spanning our three complementary perspectives: conceptual, theoretical, empirical.