A mobile robot tracking using Kalman filter-based Gaussian Process in wireless sensor networks

Abstract

RSSI-based localization has a variety of possible applications, and the environment to obtain the required information is well-constructed in these days due to the prevalence of WiFi usage. However, it is difficult to apply this method directly to the real-world positioning, because there are several factors of uncertainty in the signal strength measurements. In this paper, it is proposed to incorporate dead-reckoning using encoder measurement only, and Kalman filter-based Gaussian Process to compensate the uncertainty. As encoder itself is not able to calibrate the accumulating error, and the measured RSSI data has a time-varying error, the defects of respective methods can be complemented by each other using Kalman filter. The performance of the proposed method is evaluated by two different simulations. The location of a mobile robot moving through the exact desired path is estimated first. Then, the result of controlling a mobile robot based on the estimated position is shown.