Distributed Consensus Computing Problem Over Heterogeneous Opportunistic Networks

This paper investigates the distributed consensus computing problem over heterogeneous opportunistic networks with heterogeneity in communication, caching and computing power. In existing work on distributed consensus, the features of heterogeneous opportunistic networks have not been fully considered. This paper combines communication theory and control theory. It first introduces the quantified cache mechanism, proves the stability of the cache queue using Lyapunov drift, and designs a system protocol suitable for heterogeneous opportunistic networks. Secondly, it proves the systemic consensus and analyzes the impacts of delay, packet loss, and topological switching period on convergence. Finally, to accelerate the protocol convergence, the authors use ADMM and gradient descent methods to optimize the weight matrix, cache blocks and control sequences.