Learning and inferring transportation routines

摘要: This paper introduces a hierarchical Markov model that can learn and infer user's daily movements through an urban community. The uses multiple levels of abstraction in order to bridge the gap between raw GPS sensor measurements high level information such as destination mode transportation. To achieve efficient inference, we apply Rao-Blackwellized particle filters at hierarchy. Locations bus stops parking lots, where user frequently changes transportation, are learned from data logs without manual labeling training data. We experimentally demonstrate how accurately detect novel behavior or errors (e.g. taking wrong bus) by explicitly modeling activities context historical Finally, discuss application called ''Opportunity Knocks'' employs our techniques help cognitively-impaired people use public transportation safely.