Adaptive multi-model monitoring of recurrent mobility patterns

Abstract

Multi-model event-triggering is a highly promising technique for efficient monitoring of processes where instead of continuous or even periodic triggering of events, communication and control is only applied after some event interrupt. In this work we investigate an adaptive multi-model monitoring technique whereby a local host that switches between the observed models informs remote hosts of these events which in turn adapt their predictions to reduce prediction error and minimize unnecessary triggering events and future model switching, thereby reducing energy consumption and communication bandwidth. The adaptive technique is examined under a real public transport bus service scenario, where local and remote hosts use a set of mobility models to track travel times and update their arrival schedules according to detected deviations, i.e., event interrupts.