Predicting network connectivity for context-aware pervasive systems with localized network availability
Yves Vanrompay, Peter Rigole, Yolande Berbers · Lirias · 2007
In pervasive computing environments the availability of network connectivity is expected to evolve in time. We propose a reactive resource scheduling mechanism that relies on a pattern learning method. Patterns in the evolution of network availability in time are predicted with Markov chains. Our scheduling mechanism is able to predict future network connectivity in terms of probabilities and makes use of historical information. The goal is to achieve an efficient use of available resources. By knowing when network connectivity can be expected, energy consumption related to network discovery is lowered. We validated our approach by applying it to a real-life scenario and illustrate initial experimental results with a prototype application.