Opportunistic Consensus SLAM: Population-Coded Map Fusion over Intermittently Connected UAV–IoT Swarms

cooperatively must merge partial maps over radio links whichappear and vanish with distance, terrain and the duty cycle ofthe IoT backhaul. Naive merging on contact counts the sameevidence repeatedly, and the resulting map is confident preciselywhere it is wrong. In this paper we propose Opportunistic ConsensusSLAM (OC-SLAM), a single fusion rule in which everyUAV holds its landmark belief as a probabilistic population code,whose natural parameters are linear in the spike counts, andmerges it with neighbours by a Metropolis-weighted consensusstep only on those links which exist at that instant. We provethat the swarm-average information is conserved exactly on everylink realisation, so that no evidence is ever double counted,and we derive in closed form the mean-square contraction factorρ(p) of the disagreement as a function of the link availability p,together with a spectral bound ρ¯ (p)<1 valid for every p ∈ (0,1].The claims are verified formally by exhaustive enumeration of all65536 link patterns in exact rational arithmetic, and a simulatedeight-UAV survey shows that with three fusion rounds at p = 0.3every UAV attains the complete map with an error of 0.63 m,against 0.58m for centralised fusion, whereas naive mergingproduces an error of 2.21m and an inconsistency (NEES) oforder 1028.