Cooperative localization for multi-airship navigation network with hybrid kernel particle filter

Cooperative Localization (CL) enables agents to enhance their self-location accuracy by leveraging additional information from neighboring nodes. In near-space airship formations, CL facilitates the autonomous maintenance of spatiotemporal references. Particle Filters (PFs) are commonly employed to address CL challenges under nonlinear and non-Gaussian conditions. However, broadcasting redundant cooperative information in large networks leads to excessive observation dimensions. Additionally, unknown disturbances and anomalous observations introduce non-Gaussian noise. These factors lead to weight degeneracy in PFs, degrading positioning accuracy. This paper proposes a novel game-theoretic CL mechanism specifically tailored for near-space airships. Our mechanism integrates a perception and strategy selection method to select collaborative nodes with higher positioning accuracy, along with a robust distributed hybrid kernel PF to mitigate non-Gaussian noise. The perception and strategy selection method is designed based on the heterogeneous investment public goods game, for which the benefit function is constructed using the Cramér–Rao lower bound to allocate more investment to nodes exhibiting superior accuracy. The distributed hybrid kernel PF optimizes the proposal distribution through adaptive important region sampling and mean-shift migration, effectively managing noise uncertainty. Simulation experiments on a two-layer network of 43 airships demonstrate that our algorithm selects optimal measurements to reduce redundancy while preserving accuracy. The results highlight improvements in positioning and timing accuracy under different noise conditions compared with other methods.