XIE er al:MANAGING RFID DATA:CHALLENGES.OPPORTUNITIES AND SOLUTIONS 1297 0.4 0.8 0.6 0.2 0.4 0.2 100 200 300 400 20 40 60 80100 Frame Size f Frame size f (a) (b) Fig.2.(a)Given the tag size n100.the relationship between the read efficiencyand the frame size f.(b)As the tag size n varies,the maximum read efficiency obtained when f*=n. is too small,then there will exist transmission collisions for efficiency in a query round approaches to I when n>5,if most of the slots,causing most tags to retransmit in the next the optimal frame size f*=n is used.If this strategy is used query round.Hence,the frame size should be dynamically for each query round,then the overall efficiency is close to 1 adjusted according to the number of tags in the current Since I is the upper bound of the overall efficiency,then the query round.Therefore,many researchers have investigated global optimal efficiency is also achieved with this strategy. into this problem.Schoute et al.analyze the impact of the The above anti-collision algorithms are conventionally de- dynamic frame size selection on the reading performance in vised towards fairly idealized settings,without sufficiently ad- slotted ALOHA protocol [15].Floerkemeier et al.propose the dressing the difficulties in real applications,e.g.,the frequent strategy of optimizing the dynamic frame size,according to the movement of tags.the signal interference among multiple Bayesian probability model [16].Vogt et al.leverage Markov RFID readers,the signal attenuation and multi-path effect, process to model the identification process,and calculate a etc.Therefore,a few research work start to focus on and series of the optimized frame sizes during the interrogation try to solve the above problems.Due to the limited scanning process[17刀]. range of a single RFID reader.multiple readers are con- Lee et al.further derive the optimal dynamic frame sizes ventionally deployed in the application scenarios.Since the to maximize the channel efficiency [18].Their research con- former research work mainly focus on resolving the collisions clusions indicate that,if the frame size used in each query among multiple tag transmissions,without considering the round is equivalent to the number of tags to be identified, signal interference problem among multiple readers,literature then the maximum efficiency can be achieved for the channel.[19,20]propose optimized activating and scheduling schemes The detail principle is described as follows:assume the current for multiple readers.In this way,multiple readers are able number of tags within the effective scanning range is n,the to collaboratively avoid the signal transmission collisions. frame size is f.As the number of tags in each slot conforms Furthermore,by utilizing the mobile RFID reader,Sheng et al. to the Binomial distribution,then,the expected number of sin-develop efficient schemes for continuous scanning operations gleton slots in this frame is E[m]=nx(1-1/f)"-1.In order defined in both spatial and temporal domains [21].Their to maximize the channel efficiency i.e.,the ratio of the basic idea is to fully utilize the information gathered in the number of singleton slots to the frame size,we get the value previous scanning operations to reduce the scanning time of of f through the extremum of':aE並=0→f*=n, the succeeding ones.The former research work mainly devise then the maximum channel efficiency is n/f*1/e.In optimized schemes to identify statically deployed tags in fairly Fig.2(a)and Fig.2(b),we provide more detail descriptions for idealized settings,without considering the impact of some the above properties.Fig.2(a)depicts the impact of the frame ubiquitous issues like path loss in the mobile environment.In size f on the channel efficiency,when the tag size n =100. regard to this problem,Xie et al.propose a probabilistic model Note that as the frame size gradually increases from 1 to 400, for RFID tag identification [22],according to the continuous the read efficiency gradually increases to the maximum value changing properties of signal attenuation.Based on this model, and then decreases.The maximum efficiency is achieved they devise optimal parameters for the tag identification, when f=100.Fig.2(b)depicts the impact of the tag size n which conforms to the EPC Gen2 standard.Moreover,in on the channel efficiency,when the optimal frame size f*=n order to execute the continuous scanning with mobile reader is used.Note that when the value of n is small,the optimal for tag identification in realistic settings,Xie et al.conduct efficiency is larger than 1,e.g.,when the tag size n =1,comprehensive experimental study on mobile RFID reading, the optimal efficiency is 100%if the frame size is set to 1.and design very efficient algorithms to maximize the time- As the value of n gradually increases,the optimal efficiency efficiency and energy-efficiency by skillfully adjusting the quickly converges to It implies that,the local optimal read readers power and moving speed [23].In order to identifyXIE et al.: MANAGING RFID DATA: CHALLENGES, OPPORTUNITIES AND SOLUTIONS 1297 0 100 200 300 400 0 0.1 0.2 0.3 0.4 Frame Size f Read Efficiency n1/f (a) 20 40 60 80 100 0 0.2 0.4 0.6 0.8 1 The maximum of read efficiency n1/f Frame size f (b) Fig. 2. (a) Given the tag size n = 100, the relationship between the read efficiency n1 f and the frame size f. (b) As the tag size n varies, the maximum read efficiency obtained when f∗ = n. is too small, then there will exist transmission collisions for most of the slots, causing most tags to retransmit in the next query round. Hence, the frame size should be dynamically adjusted according to the number of tags in the current query round. Therefore, many researchers have investigated into this problem. Schoute et al. analyze the impact of the dynamic frame size selection on the reading performance in slotted ALOHA protocol [15]. Floerkemeier et al. propose the strategy of optimizing the dynamic frame size, according to the Bayesian probability model [16]. Vogt et al. leverage Markov process to model the identification process, and calculate a series of the optimized frame sizes during the interrogation process [17]. Lee et al. further derive the optimal dynamic frame sizes to maximize the channel efficiency [18]. Their research con￾clusions indicate that, if the frame size used in each query round is equivalent to the number of tags to be identified, then the maximum efficiency can be achieved for the channel. The detail principle is described as follows: assume the current number of tags within the effective scanning range is n, the frame size is f. As the number of tags in each slot conforms to the Binomial distribution, then, the expected number of sin￾gleton slots in this frame is E[n1] = n×(1−1/f)n−1. In order to maximize the channel efficiency n1 f , i.e., the ratio of the number of singleton slots to the frame size, we get the value of f through the extremum of n1 f : ∂E[n1]/f ∂f = 0 → f ∗ = n, then the maximum channel efficiency is n1/f ∗ → 1/e. In Fig.2(a) and Fig.2(b), we provide more detail descriptions for the above properties. Fig.2(a) depicts the impact of the frame size f on the channel efficiency, when the tag size n = 100. Note that as the frame size gradually increases from 1 to 400, the read efficiency gradually increases to the maximum value 1 e and then decreases. The maximum efficiency is achieved when f = 100. Fig.2(b) depicts the impact of the tag size n on the channel efficiency, when the optimal frame size f ∗ = n is used. Note that when the value of n is small, the optimal efficiency is larger than 1 e , e.g., when the tag size n = 1, the optimal efficiency is 100% if the frame size is set to 1. As the value of n gradually increases, the optimal efficiency quickly converges to 1 e . It implies that, the local optimal read efficiency in a query round approaches to 1 e when n > 5, if the optimal frame size f ∗ = n is used. If this strategy is used for each query round, then the overall efficiency is close to 1 e . Since 1 e is the upper bound of the overall efficiency, then the global optimal efficiency is also achieved with this strategy. The above anti-collision algorithms are conventionally de￾vised towards fairly idealized settings, without sufficiently ad￾dressing the difficulties in real applications, e.g., the frequent movement of tags, the signal interference among multiple RFID readers, the signal attenuation and multi-path effect, etc. Therefore, a few research work start to focus on and try to solve the above problems. Due to the limited scanning range of a single RFID reader, multiple readers are con￾ventionally deployed in the application scenarios. Since the former research work mainly focus on resolving the collisions among multiple tag transmissions, without considering the signal interference problem among multiple readers, literature [19, 20] propose optimized activating and scheduling schemes for multiple readers. In this way, multiple readers are able to collaboratively avoid the signal transmission collisions. Furthermore, by utilizing the mobile RFID reader, Sheng et al. develop efficient schemes for continuous scanning operations defined in both spatial and temporal domains [21]. Their basic idea is to fully utilize the information gathered in the previous scanning operations to reduce the scanning time of the succeeding ones. The former research work mainly devise optimized schemes to identify statically deployed tags in fairly idealized settings, without considering the impact of some ubiquitous issues like path loss in the mobile environment. In regard to this problem, Xie et al. propose a probabilistic model for RFID tag identification [22], according to the continuous changing properties of signal attenuation. Based on this model, they devise optimal parameters for the tag identification, which conforms to the EPC Gen2 standard. Moreover, in order to execute the continuous scanning with mobile reader for tag identification in realistic settings, Xie et al. conduct comprehensive experimental study on mobile RFID reading, and design very efficient algorithms to maximize the time￾efficiency and energy-efficiency by skillfully adjusting the readers power and moving speed [23]. In order to identify