ALARM: Adversarially-learned Labels using Activity and Reward Models

Technological advances in navigation and positioning, along with expanding wireless infrastructure and remote sensing technologies, have resulted in an explosive growth of available trajectory data from a variety of moving objects, such as people, cars, ships, or animals. Traditional trajectory mining algorithms do not explain how and why the motion was generated, limiting their utility in GEOINT applications when data is unlabeled, noisy, and does not contain contextual layers. Aptima Inc. will develop a system that analyzes spatiotemporal trajectory data to produce Adversarially-learned Labels using Activity and Reward Models (ALARM). ALARM will generate semantic labels of the trajectories by inferring two primary variables that cause and thus explain motion behavior: (a) intent, which we infer as a spatial reward function of the trajectories; and (b) behavior, which we infer as the motion activity policies that targets execute to reach their goals. We achieve this by using innovative variational inference, deep parameterized representations, and adversarial learning. ALARM will significantly enhance the state of the art in learning models that explain motion patterns by developing solutions that enhance observed trajectory data with simulated tracks, filling the gaps and removing bias from input observations while efficiently scaling to a large number of targets.