Cooperative RSSI-Based Indoor Localization: B-MLE and Distributed Stochastic Approximation

In this paper, we consider the localization problem on wireless sensor networks in indoor environments, in the case when low cost ranging techniques are used. Each sensor node seeks to estimate its local map (i.e., its own position and that of the sensor nodes in its neighborhood) by collecting noisy measurements of the received signal strength indicator (RSSI) from packets sent by its neighbors. In order to take into account the effects of signal attenuation usually observed in real indoor scenarios, we use the biased log- normal shadowing model (BLNSM), which relies on the assumption that the parameters of the propagation model might differ from place to place, coupled with an existing biased-maximum likelihood estimator (B-MLE), which has been proved to be more accurate for indoor environments than the standard MLE. The goal of this paper is to improve the accuracy of the positions estimated through the B-MLE while solving the network positioning problem in a cooperative manner. With this aim, we propose a two phase algorithm. At first, the initial estimated nodes' positions are obtained from the B-MLE. Secondly, the estimated positions are refined by using an on-line distributed stochastic approximation algorithm (DSA). We present numerical results related to three experimental campaigns involving different dimensions, and different radio devices (TMote Sky and TinyOS CC2420). Our tested wireless sensor networks are issued to the standard ZigBee IEEE 802.15.4 operating at 2.4GHz. In all scenarios considered here, the localization accuracy of more than the 70% of the estimated positions are improved.