Decentralized movement pattern detection amongst mobile geosensor nodes

Abstract

Movement patterns, like flocking and converging, leading and following, are examples of high-level process knowledge derived from low-level trajectory data. Conventional techniques for the detection of movement patterns rely on centralized "omniscient" computing systems that have global access to the trajectories of mobile entities. However, in decentralized spatial information processing systems, exemplified by wireless sensor networks, individual processing units may only have access to local information about other individuals in their immediate spatial vicinity. Where the individuals in such decentralized systems are mobile, there is a need to be able to detect movement patterns using collaboration between individuals, each of which possess only partial knowledge of the global system state. This paper presents an algorithm for decentralized detection of the movement pattern flock, with applications to mobile wireless sensor networks. The algorithm's reliability is evaluated through testing on simulated trajectories emerging from unconstrained random movement and correlated random walk. © 2008 Springer-Verlag Berlin Heidelberg.