Answers are grounded in this paper and cite their sources.
Traditional received signal strength indicators (RSSI's)-based moving target localization and tracking using wireless sensor networks (WSN's) generally employs lateration/angulation techniques. Although this method is a very simple technique but it creates significant errors in localization estimations due to nonlinear relationship between RSSI and distance. The generalized regression neural network (GRNN) being a one-pass learning algorithm is well known for its ability to train quickly on sparse data sets. This paper proposes an implementation of GRNN as an alternative to this traditional RSSI-based approach, to obtain first location estimates of single target moving in 2-D in WSN, which are then further refined using Kalman filtering (KF) framework. Two algorithms namely, GRNN + KF and GRNN + unscented KF (UKF) are proposed in this paper. The GRNN is trained with the simulated RSSI values received at moving target from beacon nodes and the corresponding actual target 2-D locations. The precision of the proposed algorithms are compared against traditional RSSI-based, GRNN-based approach as well as other models in the literature such as traditional RSSI + KF and traditional RSSI + UKF algorithms. The proposed algorithms demonstrate superior tracking performance (tracking accuracy in the scale of few centimeters) irrespective of nonlinear system dynamics as well as environmental dynamicity.
No takes yet. Share an insight, caveat, or question.
Jondhale et al. (2018) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: