Self-localization given single perturbed measurement of anchor positions and RSSI

Abstract

We propose a new multilateration algorithm for localization of a mobile node with limited resources given inaccurate anchor positions and received signal strength indicator (RSSI) measurements. Most of the previous works assume additive Gaussian perturbations in both distance measurements and anchor positions. However, we assume log-normal distribution for the RSSI-based distance observations as we adopt the log-distance path-loss model for radio signal propagation. To achieve an accurate and computationally-efficient localization, we formulate an appropriately-weighted non-linear least-squares problem and solve it through the gradient-descent method. The calculation of the weights is a key component of the proposed algorithm and is carried out using the statistical properties of the perturbations present in the observations of both anchor positions and RSSI-induced distances. Simulation results show that the proposed algorithm outperforms a previously-proposed related algorithm with similar computational complexity in terms of both root mean square error and bias.