Cooperative node localization schemes that employ nonlinear data reduction often deliver higher network node position accuracy compared to many other approaches. Other advantages of such algorithms are that they require only a minimum number of anchor nodes (if we require absolute locations) and that they can be applied under both range-based and range-free conditions. This article presents a novel cooperative node localization scheme, applying an efficient neural network nonlinear projection method called Curvilinear Component Analysis (CCA). A thorough comparative performance study of the proposed scheme in different mission-critical operational network scenarios is conducted. Compared with another leading cooperative node localization algorithm, MDS-MAP, which employs Multi-Dimensional Scaling (MDS), the proposed CCA-MAP approach significantly improves position estimate accuracy in many of the scenarios. We also propose a new local edge model for range-free distance matrix approximation that considerably enhances the performance for both MDS-MAP and CCA-MAP in certain irregular network configurations which are very challenging for node positioning.

Additional Metadata
Keywords Curvilinear component analysis, Localization, Multi-dimensional scaling, Nonlinear mapping, Simulations
Persistent URL dx.doi.org/10.1145/1464420.1464421
Journal ACM Transactions on Sensor Networks
Citation
Li, L. (Li), & Kunz, T. (2009). Cooperative node localization using nonlinear data projection. ACM Transactions on Sensor Networks, 5(1). doi:10.1145/1464420.1464421