Centroid, Kalman Filtering and Neural Network based Approach for Mobile Target

In recent years, the knowledge of location is of prime importance for success of many location-based services (LBS) applications such as indoor navigation, emergency management, elderly tracking, etc. There are many popular approaches in the literature which utilize received signal strengths (RSSs) in wireless sensor network (WSN) to track the moving target. Many times for the applications wherein localization accuracy to the scale of few centimetres is required, the application of one localization approach is not sufficient. The important reason is the highly fluctuating nature of RSS’s, and dynamicity in target motion and surrounding environment. Although generalized regression neural network (GRNN) has very nice estimation ability, we believe that the addition of the concept centroid to the GRNN architecture and fusing it with kalman filter (KF) can further improve its target localization performance. This paper proposes two robust range-free RSS based target localization algorithms namely, CGRNN+KF, and CGRNN+UKF, obtained by the fusion of fundamentals of centroid, GRNN, and KF. From the results obtained, it is observed that the proposed CGRNN+KF, and CGRNN+UKF algorithms yield positioning accuracy of few centimetres and effectively deal with the highly fluctuating nature of RSS’s, and dynamicity in target motion and surrounding environment.