Manifold learning algorithms for localization in wireless sensor networks

If a dense network of static wireless sensors is deployed to measure a time-varying isotropic random field, then sensor data itself, rather than range measurements using specialized hardware, can be used to estimate a map of sensor locations. Furthermore, distributed and scalable sensor localization algorithms can be derived. We apply the manifold learning algorithms, Isomap, locally linear embedding (LLE), and Hessian LLE (HLLE). The HLLE-based estimator demonstrates the best bias and variance performance, but may not be robust for all random sensor deployments.