Automatic mobility status estimation in wireless self-organised networks

Abstract

Wireless sensor networks and ubiquitous computing are rapidly increasing in popularity and diversity. For many applications of these systems the mobility status of devices forms part of the operating context on which self-organisation is based. This paper describes a novel technique by which wireless devices such as sensor nodes can deduce their own mobility status, based on analysis of patterns in their local neighbourhood. The Self-Detection of Mobility Status algorithm (SDMS) uses a reinforcement learning inspired mechanism to combine the indications from five mobility metrics. For many systems in which a neighbour table is maintained through regular status messages or other interaction, the technique incurs no additional communication overhead. The technique does not require that nodes have additional information such as absolute or relative locations, or neighbourhood node density. The work considers systems with heterogeneous time-variant mobility models, in which a subset of nodes follows a random walk mobility model, another subset follows a random waypoint mobility model (i.e. has intermittent movement), some nodes have group mobility and there is a static subset. We simulate these heterogeneous mobility systems and evaluate the performance of SDMS against a number of metrics and in a wide variety of system configurations.