 Research
 Open Access
On the optimal configuration of framed slotted ALOHA
 YoungBeom Kim^{1}Email author
https://doi.org/10.1186/s1363801607016
© Kim. 2016
 Received: 18 July 2014
 Accepted: 18 August 2016
 Published: 1 September 2016
Abstract
The primary task of RFID readers is to identify multiple objects as quickly and reliably as possible with minimal power consumption and computation. One issue addressed in this study is the manner in which RFID systems adopting the framed slotted ALOHA (FSA) protocol should be configured so that the minimum identification delay can be achieved with a preassigned confidence level. We first formalize two optimization problems associated with the configuration of passive RFID tag identification processes, i.e., the determination of two key parameter values—termination time and frame size. Next, by considering a comparative RFID system, which well approximates the original system, we establish closedform formulas for the aforementioned parameters. Furthermore, through asymptotic analysis, we investigate the asymptotic behavior of the optimal parameter values as functions of number of tags and confidence level. Through numerical comparisons, we verify the effectiveness of the proposed solutions.
Keywords
 RFID
 Framed slotted ALOHA
 Passive RFID tags
 Anticollision protocols
1 Introduction
The radio frequency identification (RFID) market was worth US$8 billion in 2013, up from US$7 billion in 2012, and grew to US$9 billion in 2014. The IDTechEx forecast is that it will be worth approximately US$30 billion in 2024 [1]. In particular, rapid growth is being witnessed in such applications as apparel tagging in the retail sector, RFIDbased ticketing in public transport systems, and tagging of animals (such as pets and livestock). It is noteworthy that recent renewed interest in RFID suggests that lowcost passive RFID tags will play an important role in the Internet of Things (IoT)—the network of interconnected things/devices which are embedded with sensors, software, network connectivity and necessary electronics that enables them to collect and exchange data [2–5]. In order for objects to interface with the existing Internet, they must have some means to connect to it: it is envisioned that RFID technology, which provides efficient wireless object identification, will bridge this gap between physical world and the virtual world, although other means are also being used, including oldfashioned barcodes, QR (quick response) codes, and wireless communication technologies such as Bluetooth and WiFi [2–5]. It is worth noting that the primary source of the RFID market growth is passive RFID tags, primarily because of their lowcost, zero maintenance requirements, and ease of largescale deployment [1, 2, 5–7].
Collisions resulting from simultaneous tag responses are one of the key issues in RFID systems. Collisions lead to losses in bandwidth and energy, and increase tag identification delays. Hence, anticollision algorithms are required to minimize these collisions. The task of designing anticollision protocols is rendered more challenging because the tags are required to be simple, cheap, and sufficiently small [7–9]. Among several anticollision protocols proposed thus far, the framed slotted ALOHA (FSA) protocol is the most widely used owing to its performance and simple implementation [6–11]. For example, RFID tags of Type A of ISO/IEC 180006 and 13.56 MHz ISM band EPC Class 1 use the FSA protocol, and Type B of ISO/IEC 180006 and 900 MHz EPC Class 0 use the binary search protocol [6, 10, 12].
Basically, RFID readers are required to identify multiple objects as quickly and reliably as possible, with minimal power consumption and computation. These performance requirements can be quantified in terms of identification delay and confidence level. One issue addressed in this study is the manner in which RFID systems adopting the FSA protocol should be configured so that the minimum identification delay can be achieved, while all the tags in the reader’s interrogation zone can be identified with a probability no lower than a given confidence level α (this probability is hereafter referred to as the probability of successful identification). For instance, a confidence level of α=0.99 implies that, if the tag reading procedure is carried out 100 times, then the reader is required to successfully identify all the tags in its interrogation zone, at least in 99 cases, on average.
Formally, we consider two problems associated with the configuration of tag identification processes. The first problem is to determine the so called termination time, T _{ α }, for given number of tags N, frame size L, and confidence level α. We need to find an appropriate value of T _{ α } such that when the process terminates at the end of the \(T_{\alpha }^{\ th}\) frame, the probability of successful identification is no lower than a given confidence level α. Once T _{ α } is determined for a given frame size L, the identification delay τ _{ α } can be computed as τ _{ α }=L×T _{ α }. Because T _{ α } and τ _{ α } are functions of L, the second problem can be defined as determining the optimal frame size L _{ opt } that minimizes τ _{ α }.
To the best of our knowledge, no closedform solution to the two foregoing problems is currently available, primarily because closedform expressions for parameters T _{ α } and L _{ opt } do not exist, and they can be computed numerically only for small values of N. In this study, by considering a comparative (imaginary) RFID system which wellapproximates the original system, and for which closedform expressions for the aforementioned parameters are available, we establish closedform formulas for the termination time, identification delay, and optimal frame sizes. Furthermore, through asymptotic analysis, we derive concise formulas and show how the optimal parameter values roughly behave as functions of number of tags N and confidence level α.
In reality, the number of tags N is unknown to the reader and the reader should estimate it using an appropriate tag estimation function, utilizing the (aforementioned) collected statistics at the end of each frame. The tag estimate, namely \(\widetilde N\), is again utilized in computing the appropriate frame sizes to be used for the subsequent frames. In this scenario, frame size L is adjusted according to \(\widetilde N\) at appropriate time instances (e.g., on a framebyframe basis) during the identification process. Consequently, the socalled dynamic FSA (DFSA) protocol [6, 7, 15–17] is more realistic than FSA. However, in this work, we assume that the frame size is fixed throughout the identification process in order to facilitate our discussion, because our interest is focused primarily on the optimal parameter configuration of the FSA protocol for a given N, i.e., we consider the situation after N or \(\widetilde N\) is determined. In addition, we assume that each tag transmits its ID once every frame, regardless of whether the previous transmissions were successful; i.e., the muting function [6, 7] is not available. Therefore, the results herein are not restricted only to the FSA protocol.
The rest of this paper is organized as follows. Two problems associated with the RFID tag identification processes are formulated in the next section. In Section 3, we introduce important research results related to our work that were reported in previous studies, and briefly discuss the problems therein. In Section 4, our approach to the solution of the tag identification problems is introduced. In Section 5, we validate the accuracy of the practical formulas based on the proposed solutions through numerical comparisons. In Section 6, we establish concise formulas through asymptotic analysis and investigate the asymptotic behavior of the optimal parameter values as functions of number of tags N and confidence level α. Finally, Section 7 concludes the paper.
2 Tag identification problems
In this section, we formalize two problems related to the optimal configuration of the FSA protocol for passive RFID tag identification. We consider only the cases where the tags are not informed by the reader about the outcome of each reading frame. Therefore, in the case of multiple reading frames, each tag transmits its ID once every frame regardless of whether the previous transmissions were successful.
We first observe that \({\mathcal {P}}_{id}^{t}\) is a nondecreasing function of frame size L; i.e., larger FSA frame sizes imply less competition among tags for time slots, but this has a drawback of prolonged identification delay. On the other hand, as we execute more FSA frames for tag identification, the probability of successful identification increases because a larger number of FSA frames implies more chances (time slots) of ID transmission for tags, i.e., \({\mathcal {P}}_{id}^{t} \le {\mathcal {P}}_{id}^{t+1}\). This inequality also follows from the fact that \({\mathcal {E}}_{t} \subseteq {\mathcal {E}}_{t+1}\), t=1,2,…. In other words, the probability of successful identification increases as the tag ID reading process proceeds from frame to frame.
The primary objective of tag readers is to identify all the tags present in the interrogation zone with a probability greater than or equal to a preassigned confidence level α in the shortest time. Therefore, we first need to know when to terminate the reading process such that the probability of the readers failing to identify all the tags is less than 1−α. In this respect, the first problem can be stated as follows.
It is worthwhile to determine the optimal frame size L _{ opt } that minimizes τ _{ α }. The optimization problem associated with the tag identification process can be stated as follows.
FSA optimization problem (FOP) For N tags and a given confidence level α, determine the optimal frame size L _{ opt } such that the identification delay τ _{ α } is minimized.
where for a function of L, i.e., f(L), arg minL=N,N+1,…f(L) denotes the value of L maximizing f(L) among L=N,N+1,…. Further, we denote the termination time for L=L _{ opt } as \(T_{\alpha }^{*}\).
3 Related work
In this section, we discuss some representative past studies on the optimal configuration of the FSA protocol and explain some problems associated with these studies.
Normalized throughput U _{ norm } represents the utilization of FSA slots in the identification process and can be interpreted as the probability that a slot has a single transmission during an FSA frame. Normalized throughput is maximized at L=N.
Floerkemeier [18], Chen et al. [19, 20], and BuenoDelgado et al. [21] proposed using L=N as the optimal frame size. Similarly, Huang [22] suggested the formula L _{ opt }=1/(1−e ^{2 ln2/N }). On the other hand, Lee et al. [12] derived the formula L _{ opt }=1/(1−e ^{−1/N }). It should be noted that all these results have one commonality: they are obtained in terms of normalized throughput or system efficiency and are not optimal with respect to the optimization problem addressed in this work.
Frame sizes for tag estimates suggested in [13]
L  16  32  64  128  256 

Low(N)  1  10  17  51  112 
High(N)  9  27  56  129  ∞ 
The solutions for BIP and FOP are obtained as follows. Suppose there are N tags and the confidence level is given by α. For each L (L=N,N+1,…), we first compute \({\mathcal {P}}_{id}^{t} (t = 1,2, \ldots)\) using (6) and then T _{ α } and τ _{ α } are determined from (2) and (3). The optimal frame size L _{ opt } corresponds to the value of L that yields the smallest value of τ _{ α }. It is noteworthy that although MC analysis provides a means of computing exact values of L _{ opt } and T _{ α }, the computation is extremely complicated and sometimes intractable for large values of N.
4 Proposed solutions for the tag identification problems
In this section, we first derive a closedform solution for T _{ α } and then proceed to the solution of the optimization problem introduced in Section 2.
In this work, we propose using \(\widetilde {\mathcal {P}}_{id}^{t}\) as a replacement of \({\mathcal {P}}_{id}^{t}\) for the solution of BIP and FOP—hereafter, we use the notations \(\widetilde {T}_{\alpha }\) and \(\widetilde {L}_{opt}\) for the solutions of (2) and (4) when \(\widetilde {\mathcal {P}}_{id}^{t}\) is used instead of \({\mathcal {P}}_{id}^{t}\). The observation (O1) has an important implication that it secures the reliability constraint in (1) because \(\alpha \le \widetilde {\mathcal {P}}_{id}^{t} \le {\mathcal {P}}_{id}^{t}\) for \(t = \widetilde {T}_{\alpha }\), even if we use \(\widetilde {\mathcal {P}}_{id}^{t}\) instead of \({\mathcal {P}}_{id}^{t}\) in (2) and (4).
The optimal termination time \(\widetilde {T}_{\alpha }^{*}\) and the corresponding minimum identification delay \(\widetilde {\tau }_{\alpha }^{*}\) can be computed by substituting \(L = \widetilde {L}_{opt}\) in (8) and (9), respectively.
5 Numerical comparison results
To validate the accuracy of the proposed solutions, we computed termination time \(\widetilde {T}_{\alpha }\) and optimal frame sizes \(\widetilde {L}_{opt}\) using our practical formulas (8) and (10) for various values of N and α, and attempted to verify that \(\widetilde {T}_{\alpha }\) and \(\widetilde {L}_{opt}\) indeed correspond to the solutions of BIP and FOP. The exact values of T _{ α } and L _{ opt } were computed using the MC method and MATLAB. It should be noted that the computation of exact values is restricted to some small values of N, because of the arithmetic overflow errors associated with the calculation of N! (and L!) pertaining to the MC analysis.
In summary, these results indicate that our practical formulas (8) and (10) can be effectively employed as the solutions of BIP and FOP. Very rarely, the formulas (8) and (10) yield parameter values deviating from the optimal values, but still, these values can be safely adopted, producing similar identification delays with a higher level of identification reliability.
6 On the asymptotic behavior of optimal frame size and termination time
In this section, we investigate how the optimal frame size and termination time roughly behave as a function of tag number N and confidence level α, in the asymptotic regime (i.e., for large N), as the formulas (8) and (10) are somewhat complicated. Throughout we assume that N is large.
From (11) and (14), we can state the following facts on the asymptotic behavior of optimal frame size and termination time.
Corollary 1

the optimal frame size L _{ opt } depends only on N and increases linearly as a function of N;

the optimal termination time \(T_{\alpha }^{*}\) depends on N and α, and increases logarithmically either as a function of N or as a function of 1/ε.
In summary, these results suggest that the simple formulas (11) and (13) (or (14)) can be employed effectively in determining the optimal values of frame sizes and termination time associated with the FSA protocol. It is worth noting that \(\bar {\mathcal {P}}_{id}^{t}\) given by (12) slightly underestimates the probability of successful identification, thereby resulting in longer identification delays; however, the differences are not significant.
7 Conclusions
In this paper, we discussed several problems regarding the optimal configuration of the FSA protocol for passive RFID tag identification; the performance requirement is to minimize the identification delay while meeting a preassigned confidence level. We first formalized two problems concerning the optimal configuration: (1) the determination of the appropriate termination time under the constraint that the probability of successful identification should be no lower than a given confidence level α, and (2) the determination of optimal frame sizes for the minimal identification delay. By considering a comparative RFID system, which turned out to be a fine approximation of the original system, we derived closedform formulas for determining the optimal values of the key parameters. Additionally, through asymptotic analysis, we obtained concise formulas which demonstrate: (1) the optimal frame size increases linearly as a function of number of tags, and (2) the optimal termination time increases logarithmically, either as a function of number of tags or as a function of 1/ε (where ε stands for level of identification failure).
Declarations
Competing interests
The author declares that he has no competing interests.
Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.
Authors’ Affiliations
References
 IDTechEx, “RFID forecasts, players and opportunities 2014–2024”. Available at http://www.centrenationalrd.com/docs/users/le/RFIDForecasts20142024.pdf. Oct 2013.
 C Edwards, RFID tags along with the Internet of Things. Eng. Technol. Mag. 9(8) (2014).Google Scholar
 J Thrasher, A Primer on the Internet of Things & RFID. Available at http://blog.atlasrfidstore.com/internetofthingsandrfid, Jan. 2014.
 B Howard, How the Internet of Things may change the world. Available at http://news.nationalgeographic.com/news/2013/08/130830internetofthingstechnologyrfidchipssmart, Aug. 2013.
 M Belfiore, How to benefit from the Internet of Things. RFID J.11(5), 17–23 (2014).Google Scholar
 L Zhu, TS Yum, A critical survey and analysis of RFID anticollision mechanisms. IEEE Commun. Mag. 49(5), 214–221 (2011).View ArticleGoogle Scholar
 D Klair, KW Chin, R Raad, A survey and tutorial of RFID anticollision protocols. IEEE Commun. Surv. Tutor. 12(3), 400–421 (2010).View ArticleGoogle Scholar
 D Klair, KW Chin, R Raad, in Sixth IEEE International Symposium on Communications and Information Technologies. On the accuracy of tag estimation functions (SydneyAustralia, 2007).Google Scholar
 B Zhen, M Kobayashi, M Shimizui, Framed Aloha for multiple RFID objects identification. IEICETrans. Commun. E88B:, 991–999 (2005).View ArticleGoogle Scholar
 V Namboodiri, M DeSilva, K Deegala, S Ramamoorthy, An extensive study of slotted Alohabased RFID anticollision protocols. Comput. Commun. 35(16), 1955–1966 (2012).View ArticleGoogle Scholar
 K Finkenzeller, RFID handbook, Second Edition (J Wiley & Sons, New York (NY), 2003).View ArticleGoogle Scholar
 SR Lee, SD Joo, CW Lee, in The 2nd Intl. Annual Conference on Mobile and Ubiquitous Systems: Networking and Services. An enhanced dynamic framed slotted Aloha algorithm for RFID tag identification (USASan Diego, 2005).Google Scholar
 H Vogt, in International Conference on Pervasive Computing (PerCom). Efficient object identification with passive RFID tags (Zurich, Switzerland, 2002).Google Scholar
 H Vogt, in IEEE Intl. Conference on Man and Cybernetics. Multiple object identification with passive RFID tags, Hammamet, Tunisia, (2002).Google Scholar
 C Wang, M Li, J Qiao, W Wang, X Li, An advanced dynamic framedslotted ALOHA algorithm based on Bayesian estimation and probability response. Int. J. Antennas Propag. 2013: (2013).Google Scholar
 E Felemban, Performance Analysis of RFID framed slotted Aloha anticollision protocol. J. Comput. Commun. 2(1), 13–18 (2014).View ArticleGoogle Scholar
 JB Eom, TJ Lee, Accurate tag estimation for dynamic framed–slotted ALOHA in RFID systems. IEEE Commun. Lett. 14(1), 60–62 (2010).View ArticleGoogle Scholar
 C Floerkemeier, in 4th Annual Intl. Conference on Pervasive Computing and Communications Workshops. Transmission control scheme for fast RFID object identification (Pisa, Italy, 2006).Google Scholar
 W Chen, G Lin, in Proceedings of the 2006 International Conference on Wireless Networks. An efficient scheme for multiple access in a RFID system (Las VegasUSA, 2006), pp. 160–163.Google Scholar
 W Chen, G Lin, An Efficient Anticollision method for tag identification in a RFID system. IEICETrans. Commun. E89B(12), 3386–3392 (2006).View ArticleGoogle Scholar
 MV BuenoDelgado, J ValesAlonso, On the optimal framelength configuration on real passive RFID systems. J. Netw. Comput. Appl. 34(3), 864–876 (2011).View ArticleGoogle Scholar
 X Huang, Efficient Dynamic framed slotted anticollision algorithm for RFID systems. Int. J. Circuits, Syst. Signal Process. 1(2), 182–188 (2007).Google Scholar
 G Khandelwal, A Yener, K Lee, S Serbetli, in IEEE International Conference on Communications (ICCŠ06). ASAP: a MAC protocol for dense and time constrained RFID systems (IstanbulTurkey, 2006).Google Scholar
 YB Kim, Determination of optimal frame sizes in framed slotted ALOHA. Electron. Lett. 50(23), 1764–1766 (2014).View ArticleGoogle Scholar
 Z Prodanoff, Optimal frame size analysis for framed slotted ALOHA based RFID networks. Comput. Commun. 33(5), 648–653 (2010).View ArticleGoogle Scholar