RFID Tag Cardinality Estimation Algorithms

RFID systems are widely used in our daily life. One fundamental and criticalproblem in RFID system is how to read RFID tags efciently. However, an importantpart in many RFID tag reading protocols is to estimate the cardinality of RFID tags.Thus, this thesis focuses on the RFID tag cardinality estimation algorithms.This thesis can be divided into two tasks. The frst task is the improvement ofa state-of-the-art estimation algorithm. By studying the broadly used RFID readingprotocol standard, EPC C1G2(Electronic Product Code Class One Generation Two),we fnd that dependency exists among response vectors, which is neglected by someprevious works. Focusing on this dependency, we proposed a novel RFID tag cardi-nality estimation algorithm, which is both accurate and efcient, named Accurate andFast Rfd estimation, AFR. Our simulation results show that Dependency does exist inRFID tag response vector. By using this dependency, we can achieve7%improvementin reducing time cost. Thus prove the efciency of our algorithm. In addition, we canstrictly prove that our algorithm can achieve any given accuracy requirement.Based on the frst task, we studied on the second task, which is to give a unifedRFID tag estimation algorithm. This algorithm aimed at making full use of the infor-mation in the RFID tag response vector to fnd a general way in RFID tag cardinalityestimation. The key of this algorithm is the using of inclusion-exclusion principle tobuild the relationship between the response vector and the cardinality of RFID tags.Then we use maximum likelihood methods to get the most possible estimation. Ouralgorithm can not only be used in uniformly distributed response vector, but can beextend to non-uniformly distributed response vector. Our simulation showed that inuniformly distributed response vector, our result was as good as the state-of-the-artresults. In non-uniformly distributed response vector, especially geometrically dis-tributed response vector, our result exceeded compared results.