Bayesian resource discovery in infrastructure-less networks

Abstract

Resource discovery is one of the most important infrastructure-level facilities enabling the success of modern ad-hoc mobile communication networks, integrating services and capabilities provided by heterogeneous objects into the Future Internet environment. However, the highly dynamic and infrastructure-less nature of these organizations, where nodes continuously join and leave the network or change their attachment connections by moving between different coverage areas, makes resource discovery an extremely challenging tasks. In lack of specific knowledge about the availability of resources and their location over the network, flooding-based or pure probabilistic exploration approaches are the only feasible options to support search/discovery operations. By considering the high communication cost and the incomplete coverage problems respectively characterizing the aforementioned approaches we propose a novel adaptive random-walk search strategy, for resource discovery in ad-hoc networks, structured according to a selective stochastic query/response scheme where the exploration process is driven by posterior probability, using Bayesian inference and relying on the history of past discovery operations. This strategy, by also taking advantage of the scale-free properties characterizing the aforementioned ad-hoc organizations, is able to significantly contain the broadcast traffic without compromising the overall success of search operations by seamlessly accommodating to dynamic changes in resource location and network topology.