Accelerating Average Consensus in Dynamical Networks Without Compromising Accuracy, Privacy or Resilience

December 4, 2025

Many networked systems, from sensor swarms to peer-to-peer platforms, need to reach agreement while protecting individual data and remaining reliable when some agents misbehave. We study this problem for discrete-time consensus on directed communication graphs and ask: can we speed up agreement without undermining accuracy, privacy, or resilience? We propose an offline weight-design procedure that reshapes the interaction strengths in two stages. First, we design weights that respect a simple local state augmentation used to conceal each agent’s initial value from single-node observa- tions. Second, we adjust weights on subgraphs that exclude up to f potentially faulty agents, supporting a standard exclude and compare filtering rule. The optimization enforces structural and stochasticity constraints and directly targets a spectral proxy for convergence speed. Henceforth, we provide con- ditions under which (i) honest agents still converge to the intended consensus value, (ii) a structural non-observability property linked to privacy is preserved, and (iii) the objective relates to the contrac- tion of the consensus error. In simulations on synthetic networks (Erd ˝os–Rényi and Watts–Strogatz) with injected adversaries, the designed weights reduce the number of iterations to reach a prescribed accuracy by roughly 30–40% compared with the same protocol without optimization, while leaving per- iteration communication unchanged. We report offline compute costs, perform sensitivity analyses for detection thresholds, and examine failure cases.

Authors in the community:

• 

  J

  José Tiago Fernandes Moniz

  ist1110960

Supervisors of this institution:

• 

  G

  Guilherme Henrique Caçador Ramos

  ist164938

• 

  S

  Sérgio Daniel Gonçalves Melo Pequito

  ist154156

electrical-engineering-electronic-engineering-information-engineering - Electrical engineering, electronic engineering, information engineering

eng - English

Instituto Superior Técnico