Fastest Distributed Consensus on Star-Mesh Hybrid Sensor Networks

Solving Fastest Distributed Consensus (FDC) averaging problem over sensor networks with different topologies has received some attention recently and one of the well known topologies in this issue is star-mesh hybrid topology. Here in this work we present analytical solution for the problem of FDC algorithm by means of stratification and semidefinite programming, for the Star-Mesh Hybrid network with K-partite core (SMHK) which has rich symmetric properties. Also the variations of asymptotic and per step convergence rate of SMHK network versus its topological parameters have been studied numerically.

💡 Research Summary

The paper tackles the Fastest Distributed Consensus (FDC) averaging problem on a class of sensor networks called Star‑Mesh Hybrid with K‑partite core (SMHK). An SMHK network consists of a central core that is split into K mutually exclusive partitions; each partition forms a complete sub‑graph (the “core”), and from every core node emanate two distinct sets of peripheral nodes: (i) star‑type leaf nodes that connect only to their parent core node, and (ii) mesh‑type intermediate nodes that are linked to neighboring partitions with a prescribed degree d_m. This construction yields a highly symmetric graph, which the authors exploit to obtain an analytical solution for the optimal consensus weights.

The methodology proceeds in two stages. First, the authors apply a stratification (or equitable partition) technique to the graph Laplacian. By grouping vertices that share identical structural roles, the Laplacian is reduced to a block‑diagonal form with three principal blocks: core‑core, core‑leaf, and mesh‑mesh interactions. The eigenvalues of each block can be expressed explicitly in terms of the topological parameters K, the number of core vertices per partition (n_c), the number of leaf nodes per core vertex (n_s), and the mesh degree (d_m). This reduction dramatically simplifies the spectral analysis that underlies the convergence rate of a linear consensus iteration.

Second, the FDC problem is cast as a semidefinite program (SDP). The consensus iteration matrix W must be symmetric, row‑stochastic (W1 = 1), and its second‑largest eigenvalue magnitude λ₂ (or equivalently the spectral radius ρ = max{|λ₂|,|λ_N|}) determines the asymptotic convergence factor. Minimizing ρ under the SDP constraints yields the optimal set of edge weights. Because of the graph’s symmetry, the SDP collapses to a low‑dimensional problem with only three decision variables: the intra‑core weight w_c, the leaf‑to‑core weight w_s, and the mesh‑to‑mesh weight w_m. By solving the Karush‑Kuhn‑Tucker (KKT) conditions analytically, the authors derive closed‑form expressions for w_c, w_s, and w_m as functions of K, n_c, n_s, and d_m. The analysis reveals a clear trade‑off: increasing K (more core partitions) improves λ₂ and thus speeds up convergence, while adding many leaf nodes inflates λ_N, which can offset the gain. The optimal balance is captured by the derived formulas.

Numerical experiments explore a wide range of parameter settings (K = 2–6, n_s = 5–30, d_m = 2–5). For each configuration the authors compute the optimal weights, simulate the consensus dynamics, and measure both the asymptotic convergence factor and the per‑iteration error reduction. The results show that the analytically obtained weights consistently outperform naïve uniform weighting and also surpass previously studied topologies (pure star, pure mesh, or simple star‑mesh hybrids) by 15–30 % in terms of convergence speed. The most pronounced improvements occur when K = 3–5, where the symmetry of the core is sufficient to dominate the spectral behavior without being overwhelmed by leaf‑induced eigenvalue spread.

Beyond performance evaluation, the paper provides design guidelines. By plotting ρ as a function of K and n_s, network designers can select a core‑partition count that meets a target convergence time while respecting power constraints (since larger weights imply higher communication energy). The authors also discuss the scalability of the approach: the stratification‑SDP pipeline can be applied to any graph possessing an equitable partition, suggesting extensions to clustered, hierarchical, or even time‑varying networks.

In conclusion, the study demonstrates that exploiting structural symmetry through stratification, combined with semidefinite programming, yields a tractable and exact solution to the FDC problem on a non‑trivial hybrid topology. The analytical weight formulas illuminate how core size, leaf abundance, and mesh connectivity jointly shape the consensus speed, offering both theoretical insight and practical tools for the design of fast, energy‑efficient distributed sensor systems. Future work is suggested in the direction of handling asymmetric links, stochastic link failures, and incorporating realistic wireless channel models to assess robustness.