Distributed stochastic optimization in opportunistic networks: the case of optimal relay selection

Opportunistic Networking allows wireless nodes to exchange data and information of interest with peers in communication range. These nodes form a large, dynamic, multi-hop network on the fly. Challenging optimization problems arise, such as end-to-end routing, resource allocation (e.g., for buffer space and bandwidth), content placement etc., exacerbated by the lack of end-to-end connectivity. While globally optimal solutions are normally sought in network optimization, node actions and decisions in this context are inherently local. As a result, most solutions proposed rely on local heuristics without any guarantees about their convergence properties towards a desired global outcome. In this paper, we argue that the framework of Markov Chain Monte Carlo (MCMC) optimization can be applied to many problems in Opportunistic Networking, providing efficient local algorithms that provably converge to a globally optimal solution. As a case study, we use the problem of optimal relay selection for group communication (e.g., multicast), based on node contact patterns.