Two manifold learning techniques for sensor localization

Many applications of wireless sensor networks require that the sensor nodes be location-aware. Using RSS, TOA, TDOA and other range measurement techniques, we are able to get the measurements of dissimilarity or pairwise distance between different neighboring sensor nodes. In this paper we use these pairwise distances to make local maps, and then align the overlapping local maps to acquire relative global coordinates of sensor nodes. We present local tangent space alignment (LTSA)-based localization, and local MDS based localization (LMBL) schemes owing to development of manifold learning algothrims. Simulations show that LMBL and LTSA-based scheme outperform the LLE-based scheme.