A Unified Approach for Fast and Accurate Cardinality Estimation in RFID Systems

Radio Frequency IDentification (RFID) systems have rich applications in daily life. A crucial problem in RFID systems is to estimate the cardinality of RFID tags. Most exiting probabilistic RFID cardinality estimation algorithms utilize a certain pattern hidden in the response vector formed by the tag responses to make estimators. In this paper, we argue that finding the patterns is actually not necessary! In this paper, we present a novel approach to RFID cardinality estimation by making the full use of the whole response vector. We characterize the critical relationship between the number of RFID tags and the specific distribution of the bits in the response vector. In this paper, we consider two kinds of responses, i.e., empty / non-empty responses and empty / singleton / collision responses. To cater for RFID systems with a very large number of tags, we also generalize our estimation approach to handle geometrically distributed response vectors. Both rigid theoretical analysis and extensive simulations have been conducted and the conclusive results demonstrate that our approach is more accurate than existing state-of-the-art RFID estimation approaches.