Probabilistic Inference of Multi-Object Trajectories

This paper presents a novel approach of inferring the trajectories of moving objects in a library, a conference hall, or a large shopping mall, where GPS, Wi-Fi or cellular localization is not readily available or is deemed too expensive. Our solution is a distributed one whereby each object records all the other objects it encounters as well as the time of contact and such contact histories collectively enable the probabilistic inference of each object's whereabouts in the past as long as the map of the area is known.