Implementation and test of an RSSI-based indoor target localization system: Human movement effects on the accuracy

Abstract The movement of humans in wireless networks is one of major effects leading to significant received signal strength indicator (RSSI) variation. Using fluctuated RSSI on estimating the target position in the RSSI-based indoor localization system can give large error and poor decision of the system. In this paper, how the human movement affects the accuracy of an implemented indoor target localization system is explored by experiments, and a proposed simple RSSI filtering solution as the guideline solution to directly handle such a research problem is also presented. For our purpose, firstly, the RSSI-based indoor target localization system, which consists of design communication operations for measuring the RSSI in the wireless network and selected well-known localization methods (i.e. the min-max and the trilateration methods) for estimating the target position, is implemented and tested. Secondly, selected well-known filtering methods (i.e. the moving average and the exponentially weighted moving average filters) and the span thresholding filter (i.e. the proposed solution) are applied for reducing the RSSI variation and the estimated position error caused by the human movement. Our experiments have been carried out in an indoor environment. An LPC2103F microcontroller interfacing with a 2.4 GHz CC2500 RF module is developed and employed as the wireless node. Experimental results reveal that the estimated position error determined by the min-max and the trilateration methods significantly increases during the human movement, and converts according to human movement patterns and numbers of movement people. Also, the results demonstrate that by applying the moving average filter with a high window size and the exponentially weighted moving average filter with an optimal weighting factor to raw RSSI data, the estimated position error is not much improved. In contrast, the span thresholding filter gives better results and can directly cope with the human movement problem. In average, the localization error and the standard deviation decrease 11.921% and 42.086% in the case of the min-max method, and they decrease 44.535% and 87.154% in the case of the trilateration method.