MDS-Based Localization

Abstract

The authors present several network node localization methods that are based on multidimensional scaling (MDS) techniques. Four algorithms are introduced: MDS-MAP(C), MDS-MAP(P), MDS-Hybrid, and RangeQ-MDS. MDS-MAP(C) is a centralized algorithm that simply applies MDS to estimate node positions. In MDS-MAP(P), a local map is built at each node of the immediate vicinity, then these maps are merged together to form a global map. MDS-Hybrid uses MDS-MAP(C) to relatively localize Nr reference nodes. Then, an absolute localization method uses these Nr nodes as anchors to localize the rest of the network. Finally, RangeQ-MDS assumes the absence of an RSSI-distance mapping function. It uses a quantized RSSI-based distance estimation technique (called RangeQ) to achieve more precise hop distances than other range-free approaches do. While MDS-MAP(C), MDS-MAP(P), and MDSHybrid can be range-aware or range-free, RangeQ-MDS is partially range-aware.