Narrowband cooperative network localization

We consider the problem of distributed cooperative location estimation in motionless and mobile narrowband wireless sensor networks (WSNs), where network terminals (to be localized) are several hops away from the anchor nodes (with a priori known location). Such problem is inspired by practical setups where communication bandwidth is limited, as in underwater systems with narrow-band acoustic modems. A cooperative and distributed particle filter (Coop-PF) is designed, where network nodes exchange information locally, i.e. only with neighboring terminals. However, each node does not communicate particles. Instead, a combination of clustering, mixture models and parameterization of the communicated information is explored. The proposed approach improves performance compared to utilizing a fixed Gaussian mixture model and demonstrates accuracy comparable to state-of-the-art wideband cooperative localization, while achieving a two orders of magnitude reduction in communication. As a collateral dividend, a computationally-efficient version is also offered.