A deep stochastical and predictive analysis of users mobility based on Auto-Regressive processes and pairing functions

Abstract

With the proliferation of connected vehicles, new coverage technologies and colossal bandwidth availability, the quality of service and experience in mobile computing play an important role for user satisfaction (in terms of comfort, security and overall performance). Unfortunately, in mobile environments, signal degradations very often affect the perceived service quality, and predictive approaches become necessary or helpful, to handle, for example, future node locations, future network topology or future system performance. In this paper, our attention is focused on an in-depth stochastic micro-mobility analysis in terms of nodes coordinates. Many existing works focused on different approaches for realizing accurate mobility predictions. Still, none of them analyzed the way mobility should be collected and/or observed, how the granularity of mobility samples collection should be set and/or how to interpret the collected samples to derive some stochastic properties based on the mobility type (pedestrian, vehicular, etc.). The main work has been carried out by observing the characteristics of vehicular mobility, from real traces. At the same time, other environments have also been considered to compare the changes in the collected statistics. Several analyses and simulation campaigns have been carried out and proposed, verifying the effectiveness of the introduced concepts.