theory
Abstract
We study noise-tolerant coordination in anonymous population protocols whose agents interact in uniformly random pairs, maintain real-valued labels, and have no global clock. Each agent cools its updates using a gain $γ_k=(k+1)^{-p}$ computed from its own local interaction counter. First, for pairwise averaging with independent communication errors drawn from a fixed centred finite-variance distribution, we prove convergence with probability 1 to a common finite random value for every $1/2<p\le1$, and derive convergence-rate exponents under the same assumptions. Second, for one-dimensional localisation with fixed non-degenerate Gaussian displacement errors, we show that the one-sided RightAlign rule rectifies centred noise and cannot stabilise; a signed update based on the relative-position discrepancy restores convergence to the true agents locations up to one finite random translation. Third, we analyse the setting in which both communication errors and displacement-measurement errors depend on the physical distance between the interacting devices. For additive independent Gaussian errors, the two sources combine into one centred effective error whose variance is the sum of the distance-dependent component variances. The signed updates protocol remains convergent for every fixed finite geometry, and the average combined variance determines the leading common-translation fluctuation. These results distinguish cooling-compatible signed corrections from one-sided updates that turn persistent noise into drift.