Distributed Localization Algorithm Based on Statistical Uncorrelated Vectors

We have developed a new localization algorithm based on a set of uncorrelated discriminant vectors (SUV). Comparing to the centralized multidimensional localization algorithm MDS-MAP that has been widely used in wireless sensor networks, this algorithm can improve the localization accuracy and reduce the computing complexity. In this algorithm, the solving equation of the double centered matrix is simplified by coordinate transformation. Then a new double centered matrix is reconstructed using a set of uncorrelated discriminant vectors in order to reduce the localization error caused by the ranging error. The direct calculation of node coordinates, as well as the distributed and incremental localization, can be realized. Simulation results indicate that after incorporation of our algorithm, the range of the localization error variation decreases rapidly and the localization accuracy becomes more stable when the number of anchors increases. In the case of large ranging error, the localization accuracy of the new algorithm is increased by about 50.16% and 62.24% than that of the trilateration location method and the MDS-MAP method, respectively.