Estimation Points Assignment와 Power Control을 이용한 RFID Tag Floor Localization의 성능향상

Abstract

RFID Tag Floor Localization consists of a tag floor and an RFID reader with its antenna and localization algorithm. Tag Floor is a set of passive transponders which are placed on floor in grid with their ID and positions. Mobile robot carries an RFID reader with its antenna, which can recognize around passive transponders. And localization algorithm estimates the position of the mobile robot by using positions of recognized transponders. However, this system has low accuracy problem, so it cannot be used for real localization problems.Therefore, this paper proposes an efficient method to improve accuracy of RFID Tag Floor Localization with analysis of RFID Tag Floor Localization. In tag floor, tag recognition cases are limited and finite. Also, there exist recognition faces. In each face, same tag recognition pattern is shown. It means that only finite estimation points of RFID reader can exist in RFID Tag Floor Localization. So, if recognition ranges are known, Estimation Points Assignment (EPA) is possible. Estimation Points Assignment means that estimation points are assigned artificially for each recognition case to minimize error variance. So, it shows better performance than that of medium value method or mean method. Moreover, power control is an efficient method to improve accuracy of RFID Tag Floor Localization. It can be analyzed as like analysis of RFID Tag Floor Localization. By the analysis, it is easily to verify that power control can make various recognition cases and power control can divide the recognition faces. Also, a number of estimation points increases. In result, the variance of errors decreases in regardless of mean method or EPA. However, EPA can be used in power control and make better performance. Therefore, proposed algorithms, based on the analysis, can raise the accuracy of RFID Tag Floor Localization without additional costs. It can be verified in the results of the simulations and experiments.