中 華 大 學 碩 士 論 文
在無線射頻辨識系統中標籤辨識問題的研究 The study of the tag identification problem in
RFID systems
系 所 別：資訊管理學系碩士班 學號姓名：M09610003 王鈞毅 指導教授：李之中 博士
中 華 民 國 九十九 年 七 月
摘要
在本論文中主要針對無線射頻辨識 (RFID) 標籤辨識進行兩項研究工作，分別 為靜態標籤辨識 (Static tag identification)及動態標籤辨識 (Mobile tag identification)。
在靜態標籤辨識的研究中，目前以時槽式ALOHA為最常使用的辨識方法。時槽式 ALOHA主要有兩項特點。第一項特點，當標籤的數量和時槽的數量相等時，我們可 以得到最佳的辨識效率。其次，若標籤的數量非常小，時槽數量與標籤數量相等 時，辨識效率將隨著標籤數量的減少而提高。目前大部分的學者對於第二項特點並 無深入的討論。因此，在靜態標籤辨識部份，本篇論文提出一個結合標籤分群和動 態 框 架 時 槽 式 ALOHA 的 標 籤 防 碰 撞 方 法 (GB-DFSA) ， 此 方 法 除 了 使 用 時 槽 式 ALOHA的第一項特點外，更使用時槽式ALOHA的第二項特點，將全部的標籤分為許 多小群組，並依序對這些群組進行辨識。因此，相較於其他使用第一項特點的方法，
本篇論文所提出的方法-GB-DFSA，有很高的辨識效率。以GB-DFSA和其他學者提出 的方法(P-EDFSA和DFSA)比較，結果顯示，GB-DFSA的辨識效率(40%)高於其他兩 者(P-EDFSA:38% DFSA:31%)。
在目前的標籤辨識研究中，大部份的研究以靜態標籤辨識為主，只有少數研究 針對動態標籤進行探討。與靜態標籤不同，動態標籤會隨意移入、移出讀取器，
(Reader)的感測範圍(Reader field)。因此，如何有效地辨識動態標籤是我們所關心的 研究議題之一。在本篇研究中，以 ALOHA 方法為基礎，利用時槽式 ALOHA 的第一 項特點，對標籤的數量進行估計，並利用此估計值決定下一週期(Read cycle)的框架 大小，使框架大小接近標籤的數量，達到最佳的辨識效率。此外，讓框架中的時槽 在辨識開始前進行廣播，告知到達標籤(Arrival tag)目前框架內可供使用的時槽範圍。
當標籤讀取廣播訊息後，即可迅速的參與當週期的辨識，減少等待時間。因此，此篇 論文以上述兩項特點，針對動態標籤辨識提出一個基於動態時槽式 ALOH 的辨識方 法 (MT-EDFSA)。以 MT-EDFSA 和其他方法(DFSA 和 MT-DFSA) 進行比較，結果 顯示，本篇論文所提出的方法─MT-EDFSA 其辨識每一標籤的時間(Service time)低 於其他方法，而且在處理標籤辨識的能力，皆較其他兩者來得高。
關鍵字 關鍵字 關鍵字
關鍵字: 無線射頻辨識無線射頻辨識無線射頻辨識無線射頻辨識；；；；標籤防碰撞標籤防碰撞標籤防碰撞標籤防碰撞; 動態標籤動態標籤動態標籤動態標籤; DFSA; GB-DFSA; MT-EDFSA
ABSTRACT
This research performed two works on RFID tag identification – static tag identification and mobile tag identification. In the static tag identification, slotted ALOHA is the most common method to make the tag identification. Slotted ALOHA have two features. The first feature is that the optimal throughput of slotted ALOHA increased if the number of tags and the number of slots are equal. The second feature is that if the number of tags is small and the frame size is equal to the number of tags, the throughput increases as the number of tags decreases. However, most researchers have made little effort to explore the second feature to increase the throughputs of Slotted ALOHA. This research proposes a grouping based dynamic framed slotted ALOHA anti-collision method with fine groups (GB-DFSA) to fully utilize the two features to perform tag identification. This research compares the throughput of GB-DFSA with those of the other two methods – EDFSA with partition and DFSA. The results show that the throughput of GB-DFSA is 40% which is higher than EDFSA (38%) with partition and DFSA (31%).
In the mobile tag identification, mobile tags arrive at or leave the reader’s field randomly. This research proposed an ALOHA based mobile tag anti-collision algorithm - Enhanced dynamic framed slotted ALOHA for mobile tags (MT-EDFSA). MT-EDFSA estimates the number of unidentified tags as the frame size to raise the throughput of tag identification. MT-EDFSA broadcasts the identification information to arrival tags at the beginning. Arrival tags use this identification information to select a slot randomly and respond its EPC code in the current read cycle. This research compares the service time of MT-EDFSA with those of the other methods – DFSA and MT-DFSA. The results show that the service time of the proposed method is lower than those of other methods under different arrival rates.
Keywords: RFID; anti-collision; mobile tag; DFSA; GB-DFSA; MT-EDFSA
致謝
在研究所的這段時間裡，我學習到許多老師所授予的專業知識與待人接物的技 巧，使我在就讀的期間非常充實。本篇論文的完成，首先要感謝的是我的指導老師李 之中博士，在研究過程中遇到的問題，老師總是適時給予我建議與協助，使我的研究 工作能順利進行並完成此篇論文。此外我要特別感謝龍華科技大學資訊管理學系李銘 城老師，給予我不少實驗的指導與建議，使論文的實驗能順利進行。以及感謝口試委 員柯志坤博士及蔡耀弘博士，給予我不少研究與論文撰寫上的建議，使我的論文能夠 更完善。
很高興能夠在研究所的日子中有學駿、峻維、昇岳、遜彰、永昊以及同一間商業 智慧實驗室的佩伶和麗麗這幾位學弟妹的陪伴。因為有你們的存在，使我的研究生活 充滿活力與朝氣，真的很謝謝大家。
最後，我要感謝父母的包容與支持，答應我自願延畢這麼任性的請求，使我能完 成符合自己期望的論文。因此，我將此篇論文獻給我的家人，感謝你們在我背後的支 持與鼓勵。
TABLE OF CONTENTS
ABSTRACT (Chinese)... I ABSTRACT ...II Acknowledgement ... III TABLE OF CONTENTS ... IV FIGURE LIST ... VI
Chapter 1 Introduction ... 1
1.1. Background... 1
1.2. Tag identification ... 1
1.3. Motivation ... 2
1.3.1. Static tag identification ... 2
1.3.2. Mobile tag identification ... 3
1.4. Overview of the dissertation... 4
Chapter 2 Related works ... 5
2.1. Static tag identification ... 5
2.2. Mobile tag identification ... 6
Chapter 3 Anti-collision algorithm for static tags identification... 9
3.1. Grouping concept ... 9
3.2. Grouping based dynamic framed slotted ALOHA ... 10
3.3. Performance evaluation ... 13
3.3.1. The effects of the number of tags on throughput... 13
3.3.2. The effects of the frame size on throughput ... 14
Chapter 4 Tag identification method for mobile tags... 17
4.1. System model ... 17
4.2. Proposed mobile tag identification method ... 18
4.2.1. The concepts of mobile tag identification ... 18
4.2.2. Enhanced dynamic framed slotted ALOHA for mobile tags ... 20
4.3. Performance evaluation ... 23
4.3.1. The effects of arrival rate on service time ... 24
Chapter 5 Conclusions ... 26
References ... 27
Appendix 1 – Java code for GB-DFSA ... 29
Appendix 2 – Java code for MT-DFSA ... 34
Appendix 3 – Java code for MT-EDFSA ... 40
FIGURE LIST
FIGURE 1THE MAXIMUM THROUGHPUT AS A FUNCTION OF N.[10] ... 10
FIGURE 2THE ALGORITHM OF GB-DFSA... 12
FIGURE 3THE EFFECTS OF THE NUMBER OF TAGS ON THROUGHPUT. ... 14
FIGURE 4THE EFFECTS OF THE FRAME SIZE ON THROUGHPUT IN GB-DFSA... 15
FIGURE 5THE EFFECTS OF THE FRAME SIZE ON THROUGHPUT IN P-EDFSA... 16
FIGURE 6THE SYSTEM MODEL IN OUR RESEARCH. ... 17
FIGURE 7THE FRAME STRUCTURE OF MT-EDFSA... 20
FIGURE 8THE EFFECTS OF ARRIVAL RATE ON SERVICE TIME (ARRIVAL RATE:0.1 TO 0.5). ... 24
FIGURE 9THE EFFECTS OF ARRIVAL RATE ON SERVICE TIME (ARRIVAL RATE:0.1 TO 0.3). ... 24
FIGURE 10 THE EFFECTS OF ARRIVAL RATE ON SERVICE TIME (ARRIVAL RATE:0.3 TO 0.5). . 25
Chapter 1 Introduction
1.1. Background
Radio Frequency Identification (RFID) is an emerging technology that applies to object identification and tracking in the supply chains of businesses. Traditionally, the most popular object identification technique was through barcodes [1]. However, barcode technology has some drawbacks, such as line-in-sight and contact. Differing from barcode technology, RFID technology realizes fast and reliable identification without requiring physical sight or touching between readers and tags. For this reason, many scholars are starting to do researches in the RFID field in order to advance the RFID technology.
1.2. Tag identification
In RFID systems, one of the issues in the research community is how to identify efficiently the tags in a RFID reader field [2]. In a process of RFID tag identification, a reader first sends a request signal to tags in its field and then the tags, which received the request signal, respond by sending their EPC code to the reader immediately. Therefore, some of these tags may respond with their EPC code at the same time. Unfortunately, a reader cannot identify more than one tag at a time. When two or more tags respond to the reader’s request simultaneously, then signal is jamming and tag collision occurrs. To deal with tag collisions, many researchers have designed their anti-collision methods to resist tag collision. These methods were classified using two categories: tree-based methods and ALOHA-based methods. In tree-based methods, researchers structured the identification actions as a tree such as a query tree or a binary tree. The best advantage of the tree-based methods is an identification rate close to 100%. On the other hand, the drawback of the tree-based methods is “identify delay,” that is, if the reader must identify huge tags, tree-based methods take a long time to identify tags.
In ALOHA-based methods, the transmission period is divided into continuous time slots.
After the reader sends a request signal to tags in its field, a tag, which received the request signal randomly, chooses a slot in the transmission period to respond with its ID to the reader. The advantage of the ALOHA-based methods is they don’t have “identify delay.”
But the drawback of the ALOHA-based methods is “tag starvation,” that is, due to the randomized behavior of a tag in selecting a time slot to transmit its ID; the tag may not
occupy an appropriate time slot to successfully transmit its ID in the identification. The reader, therefore, does not identify the tag.
1.3. Motivation
How to identify RFID tags efficiently is an important issue in RFID systems.
However, a tag collision problem may occur during the reader identifies the tags. Tag collision means two or more tags respond their EPC code to the reader at the same time.
Therefore, the reader cannot identify the EPC code of each tag in these two or more tags.
To solve the tag collision problem, scholars designed many methods, called anti-collision methods or tag identification methods. In this research, there are two research issues to be explored on RFID tag identification. The first issue is that how to identify static tags in the reader field, and the second issue is that how to identify mobile tags in the reader field. The static tag means that the locations of tags are fixed during the reader identified the tags in the reader’s field, while the mobile tag means that the locations of tags are changed, that is, tags may move into or out the reader’s field, during the reader identified the tags.
1.3.1. Static tag identification
In the static tag identification, the ALOHA-based methods are one kind of methods which were most common used to solve the tag collision problem. In the ALOHA-based methods, dynamic framed slotted ALOHA (DFSA) is often utilized as a schematic. Many ALOHA-based methods [2-3,5-6,9-13] are variants of DFSA. DFSA has two features. The first feature is that DFSA reaches the highest throughput if the frame size is equal to the number of identified tags in the reader field. The second feature is that the throughput increases as the number of identified tags decreases when the frame size is equal to the number of identified tags. Most of ALOHA-based works focus on how to exploit the first feature; relatively few works try to exploit the second feature. To exploit the first feature, some ALOHA-based methods first estimate the number of tags in the reader field and then use the estimated number of tags as the frame size. To utilize the second feature, Lee [9]
divided the identified tags into several groups with few tags, and then the reader identifies the tags group by group. The throughput of dividing tags into some groups is higher than that of viewing tags as a single group. However, Lee [9] did not discuss the influence of the tag group sizes on the throughput. While few works have paid attention to exploiting the second feature, this research believes that the second feature can be further exploited.
According to the above observation, how to divide tags into small and almost equal size
groups is critical to exploiting the second feature. Therefore, this research use DFSA as a schematic to design an anti-collision method that uses tag grouping, and call this method
grouping based dynamic framed slotted ALOHA with fine group (GB-DFSA).
1.3.2. Mobile tag identification
The tag identification can classify into two categories: the static tag identification and the mobile tag identification. In the static tag identification, all unidentified tags which are in the reader field stay in the reader field until the reader identifies the tags. On the other hand, in the mobile tag identification, the tags may randomly move into or out the reader field.
There are many researches which have proposed methods on how to make the static tag identification [4-13]. These methods are classified as ALOHA- based methods and tree-based methods. In ALOHA-based methods, the transmission period is divided into continuous time slots. After the reader sends a request signal to tags in its field, a tag which received the request signal randomly chooses a slot in the transmission period to respond with its ID to the reader. The advantage of the ALOHA-based methods is they don’t have
“identify delay.” But the drawback of the ALOHA-based methods is “tag starvation,” that is, due to the randomized behavior of a tag in selecting a time slot to transmit its ID; the tag may not occupy an appropriate time slot to successfully transmit its ID in the identification.
The tag, therefore, is not identified by the reader. In tree-based methods, researchers structured the identification actions as a tree such as a query tree or a binary tree. The best advantage of the tree-based methods is an identification rate close to 100%. On the other hand, the drawback of the tree-based methods is “identify delay,” that is, if the reader must identify huge tags, tree-based methods take a long time to identify tags.
While many research works have proposed lots of methods on how to deal with the static tag identification, the mobile tag identification has not received enough attentions. Only a few works reported on how to make the mobile tag identification [16-21]. In facts, the tags are mobile in many read world applications such as a ticket hold by a passenger in the MRT system at Taipei and a ticket hold by a visitor in an exhibition. Therefore, it is necessary to further explore how to make the mobile tag identification.
To make the mobile tag identification, this research presented an ALOHA-based method which is schematic as dynamic framed slotted ALOHA (DFSA) [4] and called this method
enhanced dynamic framed slotted ALOHA for mobile tags (MT-EDFSA). In MT-EDFSA,
two subjects need to be discussed. The first subject is that, in the ALOHA-based methods[4-10], tags firstly receive the frame size and then each of them randomly chooses a slot to transmit their ID to the reader. However, in the mobile tags identification, tags do not have the frame size of the current read cycle immediately at the time of the tag arrives at the reader field if the identification procedure of the mobile tag identification is the same as that of the static tag identification. Hence, these tags cannot be identified until the next read cycle and this phenomenon may increase the time of tag identifying. Thus, a new identification procedure in the mobile tag identification is needed and makes the tag receive the frame size of the current read cycle immediately is possible. The second subject is that MT-EDFSA use DFSA as the schematic. Therefore, how to estimate the number of unidentified tags of the next read cycle is critical and need to be further explored. By considering above two subjects, this research presented MT-EDFSA to perform the mobile tag identification.
1.4. Overview of the dissertation
The rest of this dissertation is organized as follow. In chapter 2 is the related works about static tags and mobile tags identification. The static tags anti-collision algorithm – GB-DFSA and mobile tags anti-collision algorithm – MT-EDFSA is explained in chapter 3 and 4, respectively. And the performance analysis and simulation results of GB-DFSA and MT-EDFSA are also given in chapter 3 and 4, respectively. Chapter 5 is the conclusion of this dissertation.
Chapter 2 Related works
How to design an effective and efficient anti-collision method to solve the tag collision problem in RFID systems is an important issue. In the research community, researchers have proposed many anti-collision methods to solve the tag collision problem.
Here, related works about the static tags identification and the mobile tags identification are introduced.
2.1. Static tag identification
One type of anti-collision methods to solve the tag collision problem of static tags identification is called ALOHA-based, which are based on framed slotted ALOHA (FSA) [3]. In FSA, a frame is organized as a fixed number of time slots. A reader uses a frame to identify tags is called cycle. A process of identifying tags may be organized as several cycles. In an identifying tags process, a tag which is identified in the current cycle does not need to be identified again in the incoming cycles. Therefore, the number of unidentified tags is decreased cycle by cycle. However, since the frame size is fixed, FSA cannot simultaneously decrease the frame size as the number of unidentified tags decreases, and therefore identification delay is generated. To reduce identification delay in FSA, Schoute [4] proposed Dynamic Frame Length ALOHA (DFSA). DFSA counts the number of collision slots on the current cycle, nc, and then estimates the number of unidentified tags as 2.39*nc. DFSA then sets the frame size of the next cycle to 2.39*nc. Therefore, DFSA keeps the identification delay of DFSA to less than that of FSA.
In DFSA, the variation of the frame size among cycles during identification is distinguished as two phases – the growing phase and the sinking phase. In the growing phase, the frame size follows 2.39*nc to increase until the frame size approaches the number of unidentified tags. When the frame size approaches the number of unidentified tags, the throughput of tag identification may reach the optimal level. In the sinking phase, accompanying the identifying of tags, the number of unidentified tags also decreases and the frame size decreases simultaneously until all tags are identified. In fact, the contribution of the growing phase for identifying tags is limited. For this reason, some research [10-12] focuses on how to shorten the growing phase, and let identification enter the sinking phase as soon as possible. These research works proposed that how efficiently and precisely estimate the number of tags. In [5, 6], Park et al. proposed Framed Slotted
ALOHA with robust Estimation and Binary Selection (EB-FSA) to estimate the number of tags and used this number to determine frame size. When a tag collision occurred, EB-DFSA utilized binary selection to deal with collided tags. In [7], Cha et al. proposed Novel anti-collision algorithms for fast object identification. This algorithm used two estimation functions to compute the number of tags and the frame size computed by the estimated value of tags.
Wang et al. [8] used a two-functioned algorithm to estimate the number of tags in their proposed anti-collision method. This method consists of two estimation functions.
The first function estimates tag population using collision and readable slots1. The second function is used when the wrong weight in the first function is below a confidence level.
After estimating tags, this algorithm also modifies the frame size to achieve optimal performance.
Except for estimation frame size or number of tags, modifying the number of respond tags can also achieve fast identification with less collision occurrence. In [9], Lee and Lee proposed an enhanced dynamic framed slotted ALOHA anti-collision method (EDFSA).
This method consists of two phases, the first to estimate the number of unidentified tags and the second to estimate the optimal frame size in each cycle. The most related work with this paper is Shin and Kim’s work. In [10-11] Shin and Kim [10-11] expand an algorithm to portion respond tags into several tag sets; when the first set is identified, other tag sets can use the information obtained from the first set (the number of collision slots) to modify the number of tags in their groups. Shin and Kim, however, did not discuss the influence of the tag set size in term of throughput.
2.2. Mobile tag identification
In the RFID tag identification research field, most of researches concentrate the static tags identification [4-15]. In the static tag identification, all tags stay in the reader field during the reader makes the tag identification. In contrasts, in the mobile tags identification, all tags have mobility and may randomly arrive at of leave the reader field [16-21]. The mobile tags have three states – arriving, staying and leaving [16-21]. The tag is in the arriving state if the tag arrives at the reader’s field and is not identified. The tag is in the staying state if the tag is identified by the reader and the tag stays in the reader field. The tag is in the leaving state if the tag is identified and already leaves the reader field. To
1 The number of tags = 2*collision slots + 1.
confirm
In the mobile tag identification, the tags in the arriving state or in the staying state need to be identified. The reader identifies the arriving tags to have their IDs, while the reader identifies the staying tags to confirm the tags still stay in the reader field or already leaves. To solve the collision caused by the slot competition between the arriving tags and the staying tags, some scholars proposed their tree-based anti-collision methods [16-21].
In [17, 18], Myung and Lee proposed Adaptive binary splitting (ABS) algorithm. In ABS algorithm, tags are depending on progressed-slot counter (PSC) and allocated-slot counter (ASC) to transmit their id. PSC represents the number of slots passed in a frame and ASC is the sequence that a tag can access to the channel to transmission. Tags are transmitting their id when PSC equaled to ASC and the read cycle terminated when PSC large than TSC (TSC records the last slot of the frame). When the current read cycle is finished (PSC >
TSC), each tag gets unique ASC and use to transmit in the next read cycle. When the set of tags in the current read cycle equaled to the former, tags can transmit directly without collision. If there are some tags access into the current read cycle, those arriving tags are randomly choosing ASC to transmission.
Even through ABS algorithm solve the identification of mobile tags, but it has another problem – higher probability of tag collision. In ABS algorithm, arriving tags and staying tags are composite ASC in the same read cycle. Therefore, tag collision occurred when the ASC value in arriving tag and staying tag are equaled. To reduce the collision caused by staying tags and arriving tags, some scholar use blocking mechanism to restrict arriving tags’ response [19-20]. In [19, 20], Lai and Lin proposed a Pair-Resolution Blocking (PRB), which based on ABS algorithm and use blocking mechanism to divide arriving tags and staying tags into two different phases to respond. In PRB, tag’s identification divided into two phases. The first phase use rRID and tRID to distinguish between staying tag (rRID = tRID) and arriving tag (rRID ≠ tRID). After distinguishing, staying tags use their ASC to respond, and arriving tags sets their tRID to the reader’s rRID and respond in the second phase. In addition, PRB use a pair resolution scheme to identify one pair of tags.
According as this feature, the throughput of PRB is better than ABS. Similar with PRB, Yeh, Lo, Li and Winata proposed an Adaptive n-Resolution (AnR) [21], which also uses the blocking mechanism to restrict arriving tags’ response. Such like PRB, the tag’s identification in AnR divides into two phases. The first phase is use for identify staying tags and the second phase is use to identify arriving tags. In addition, AnR use
n-Resolution mechanism to cut through the time of staying tags identification. Therefore, the throughput of AnR might out perform than PRB.
Chapter 3 Anti-collision algorithm for static tags identification
In this chapter, the proposed stable tag's anti-collision method, that is, Grouping based Dynamic Framed Slotted ALOHA with fine groups (GB-DFSA) is introduced, The grouping concept and the algorithm of GB-GDSA is given first and then and the algorithm of GB-GDSA is given. Finally, the performance evaluation of GB-DFSA is reported.
3.1. Grouping concept
In slotted ALOHA, each tag occupied a slot to transmit its ID to a reader. Slotted ALOHA reached the optimal throughput2 in the following two conditions [4].
•
If the number of tags and the number of slots are unlimited and the number of slots is equal to the number of tags, then the optimal throughput of slotted ALOHA is about 36.8%.•
If the number of tags is small and the frame size is equal to the number of tags, the throughput might exceed to 36.8%.Similar with slotted ALOHA, the conditions for framed slotted ALOHA to reach its optimal throughput are the same as those of slotted ALOHA. Here, these two features are stated by using Figure 1[10]. In Figure 1, when the frame size is small, (the number of slots
< 15) and equaled to the number of unidentified tags, the throughput is greater than 38%.
In another situation, if the frame size and the number of unidentified tags are unlimited and equaled, the throughput is in the range of 36.8% to 37%.
2 Throughput = the number of identified tags/total slots used.
Figure 1 The maximum throughput as a function of n. [10]
In reality, a reader is unlikely to be aware of the number of unidentified tags before the identification process proceeds. To get the optimal throughput in the identification process, following condition 1, the frame size is equal or, at least, very close to the number of tags.
If the number of tags is unlimited, the optimal throughput is still about 36.8%.
Fortunately, following condition 2, when the number of tags is small and the frame size is equal to the number of tags, the throughput might exceed 36.8%. In fact, when both the frame size and the number of tags are 4, the throughput is about 42.2%. Therefore, if all tags in the reader field are distinguished into several small groups, and then identify tags through DFSA group by group instead of viewing all tags as a single group and, then, identifying the single group tags through DFSA, the throughput of distinguishing all tags in the field into several small groups is higher than that of viewing all tags as a single group. When the tags are divided into several small groups, the size of groups is critical to make the GB-DFSA reaches the best throughput.
Now that the grouping concept of GB-DFSA has been described, a detailed description of the method of GB-DFSA is followed with.
3.2. Grouping based dynamic framed slotted ALOHA
The tag identification process of grouping based dynamic framed slotted ALOHA with fine group (GB-DFSA) consists of two phases; the first phase is the grouping phase and the second phase is the identification phase. In the grouping phase, the reader estimates the number of tag groups it needs through a series of probes. This search denotess the number
of tag groups as s. After the number of the groups, s, is estimated, GB-DFSA enters the identification phase, that is, the second phase. In the identification phase GB-DFSA first distinguishes all tags into s groups. GB-DFSA then uses DFSA to identify the tags group by group until all tags are identified.
The algorithm of GB-DFSA is shown in Figure 2. In Figure 2, algorithm GB-DFSA has one argument – tags. Argument tags represent a set of tags which are ready to be identified by GB-DFSA. Algorithm GB-DFSA also has six variables. Let f denote the frame size, s represent the number of groups, pth be the threshold, i be the number of probing iterations in grouping phase, m be fraction factor, and tagsg represent tags of a tag group. As mentioned in the above paragraph, the grouping phase of GB-DFSA is organized into a series of probes. GB-DFSA utilizes a loop and Function FSA() to proceed with this series of probes. In the first probing, GB-DFSA views all tags as a single group and identified by using framed slotted ALOHA with frame size f. Function FSA() has two arguments, that is, the tag set, tagsg, in which tags need to identified and the frame size f of this probing. After this probing is done, Function FSA() returns the number of collision slots, denotes nc, and then checks whether Predicate (2.39*nc > pth) is true or not. If Predicate (2.39*nc > pth) is true, GB-DFSA adds 1 to the number of probing iterations, i, and set the number of group to mi-1 where m is fraction factor. GB-DFSA then chooses one of these mi-1 groups to perform the next probing iteration. If Predicate (2.39*nc > pth,) is false, the loop is ended and GB-DFSA enters the identification phase with mi-1 groups.
In the identification phase, GB-DFSA uses DFSA to identify the tags group by group.
First, GB-DFSA sets the initial frame size, f, of DFSA to pth. GB-DFSA then utilizes Function DFSA() with two parameters – the set of unidentified tags, group, and initial frame size, f - to identify the first tag group. GB-DFSA stores the identified tags in tag set identified tags. In the other groups’ identification, GB-DFSA set the initial frame size of DFSA to the average tag number of a group for all already identified groups. GB-DFSA uses a loop and Function DFSA() to identify the tags group by group. DFSA() has two arguments, the unidentified group, group, and the initial frame size, f. Function DFSA() returns the set of tags in which tags have been identified. GB-DFSA adds these identified tags to Set identified tags. The loop is ended when the identification of all groups are done.
After the two phases of GB-DFSA are completed, GB-DFSA returns Set identified tags and end the GB-DFSA.
Figure 2 The algorithm of GB-DFSA
3.3. Performance evaluation
In this section, the experiments were performed to evaluate the performance of GB-DFSA through simulations. GB-DFSA was compared with two other methods – DFSA [4] and EDFSA with partition [10, 11]. To shorten the names of EDFSA with partition, EDFSA with partition is called P-EDFSA. In order to evaluate the performance of these methods, a simulation model is developed in which only one reader was used to identify tags in the field of this reader. This research also assumed that the communication channel between the reader and these tags was ideal (i.e., error-free). Based on this model, the experiments were performed; the performance metric in the experiment was the throughput which was the ratio of the number of identified tags to the total number of slots in the identification process. The simulation model was implemented using JAVA on a PC running Windows XP Professional. Some parameters in the experiments were set as follows.
•
In GB-DFSA, the initial frame size and initial number of groups were set to 4 and 1, respectively. This search further set the threshold to 4 and the fraction factor to 4.•
In GFSA, the initial frame size was 16.•
In P-EDFSA, the initial frame size was 16 and the initial number of groups was 4.3.3.1. The effects of the number of tags on throughput
The effects of the number of tags on the throughput among GB-DFSA, DFSA, and P-EDFSA was studied in this experiment. The number of tags was varied from 100 to 2000 by step of 100 in this experiment. The experiment results are shown in Figure 3. In Figure 3, the x-axis represents the number of tags and the y-axis represents the throughput. Every point in Figure 3 represents a throughput with 95% confidence interval.
0 0.08 0.16 0.24 0.32 0.4 0.48
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Throughput
The number of tags
DFSA P-EDFSA GB-DFSA
Figure 3 The effects of the number of tags on throughput.
The first result was that GB-DFSA outperformed DFSA and P-EDFSA in terms of throughput. The throughput of GB-DFSA is near 40%, which is larger than that of P-EDFSA (38%), and that of DFSA (31%). The reason of the throughput of GB-DFSA outperformed those of the other two methods is that the strategy which divided tags into small size groups, i.e., group size near 4, worked. The second result which is worth our attention is that when the number of tags and the number of slots were unlimited and the number of slots was equal to the number of tags, the optimal throughput of slotted ALOHA was only about 36.8%. The throughput of our proposed method, GB-DFSA, in the condition of an unknown number of tags, is higher than the optimal throughput of slotted ALOHA, 36.8%. The third result is that, except when the number of tags is 100, the line of GB-DFSA is almost horizontal, and this phenomenon demonstrated that the throughput of GB-DFSA is not sensitive to the number of tags. While the worst throughput of GB-DFSA occurred with the number of tags equal to 100, in this condition the throughput of GB-DFSA still outperformed those of the other two methods, DFSA and P-EDFSA.
3.3.2. The effects of the frame size on throughput
The effect of the frame size on the throughput between GB-DFSA and P-EDFSA was studied in this experiment. The frame size was varied in 2, 4, 8 and 16 in this experiment.
The experiment results of GB-DFSA and P-EDFSA are shown in Figure 4 and Figure 5 respectively. In Figure 4 and Figure 5, the x-axis represents the frame size and the y-axis represents the throughput.
In Figure 4, the first result is that GB-DFSA with frame size 4 outperformed
GB-DFSA with frame size 2, 8, and 16 in terms of throughput. The reason of the throughput of GB-DFSA with frame size 4 outperforms those of GB-DFSA with frame size 8 and 16 is that the group size in GB-DFSA with frame size 4 is less than those in GB-DFSA with frame size 8 and 16. Following the second feature of DFSA, that is, if the number of tags is small and the frame size is equal to the number of tags, the throughput increases as the number of tags decreases, the throughput of GB-DFSA with frame size 4 outperformed those of GB-DFSA with frame size 8 and 16. Another result which needed to be discussed is that the throughput of GB-DFSA with frame size 4 outperformed that of GB-DFSA with frame size 2. The reason causing this result is that the group size in GB-DFSA with frame size 2 cannot uniformly distribute tags in each group, that is, some groups may be empty. GB-DFSA identifies these empty group don’t benefit the throughput.
The throughput of GB-DFSA with frame size 2, therefore, is less that of GB-DFSA with frame size 4.
0.35 0.36 0.37 0.38 0.39 0.4 0.41
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Throughput
The number of tags
Frame size=2 Frame size=4 Frame size=8 Frame size=16
Figure 4 The effects of the frame size on throughput in GB-DFSA.
0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4
0 200 400 600 800 1000 1200 1400 1600 1800 2000
Throughput
The number of tags
Frame size=2 Frame size=4 Frame size=8 Frame size=16
Figure 5 The effects of the frame size on throughput in P-EDFSA.
The effects of the frame size on the throughput of P-EDFSA are shown in Figure 5.
The first result is that P-EDFSA with frame size 2, 4, 8 outperforms P-EDFSA with frame size 16 in terms of throughput. The reason of this result is the second feature of DFSA worked. The second result is that the throughputs of P-EDFSA with frame size 2, 4, 8 are almost the same. It is obvious that this result is different with that of GB-DFSA, that is, GB-DFSA with frame size 4 has the best throughputs among GB-DFSA with frame size 2, 4, 8.
Chapter 4 Tag identification method for mobile tags
In this chapter the identification of mobile tags is explored. First of all, the system model and model assumptions are given and then an algorithm, Enhanced Dynamic Framed Slotted ALOHA for Mobile tags (MT-EDFSA), which is proposed to deal with mobile tag identification, is introduced. Finally, the performance evaluation of MT-EDFSA is reported.
4.1. System model
The RFID system considered in this research consists of a reader and tags, as show in Figure 6. In the system, the tags continuously arrive at the reader field and then the reader recognizes these tags by using the tag identification method. Each arrival tag has its own unique ID. In this system, any arrival tag does not leave the reader filed until the reader has identified it, that is, the reader has the unique ID of the tag. After the tag has been identified, the reader does not identify the tag again and query the tag. Since the tag has been identified, the identified tag may leave the reader field at any time. In our model, the tag has three states – the arriving state, the waiting state and the identified state. The tag is in the arriving state if the tag arrives at the reader field on the current read cycle and the reader does not identify it. The tag is in the waiting state if the tag arrives at the reader field on the previous or the earlier read cycle and the reader does not identify it. The tag is in the identified state when the tag has been identified by the reader.
Figure 6 The system model.
The proposed system model is different with some existing system models which were proposed in [17-21]. The main different between the proposed system model and the existing system models is that the proposed system model does not consider the tag may be in the staying state. In the existing system models, the reader may query the identified tags again and thus the reader need to know which tag is still in the reader field.
4.2. Proposed mobile tag identification method
In this section the mobile tag identification method which deals with tag collision among the mobile tags is specified. This method is called Enhanced Dynamic Framed Slotted ALOHA for Mobile tags (in short, MT-EDFSA). First, the concept of mobile tag identification is stated and then give the introduction of the mobile tag identification method is stated.
4.2.1. The concepts of mobile tag identification
The presented method, MT-EDFSA, is an ALOHA-based anti-collision method. In an ALOHA-based anti-collision method, the period to make identification for every tag in a reader field calls a frame. The process of using a frame to make tag identification is called a read cycle. At the every beginning of the read cycle, the reader broadcast the frame size of the current read cycle to every tag in its reader field. After the tags receive the frame size, each of them randomly chooses a slot of the frame to respond its tag ID. Any slot in which only one tag responds its tag ID is a success slot, two or more tags responds ID is a collision slot, and finally no tag responds tag ID is an idle slot. All collision tags need to be identified again in the next read cycle. The read cycles are repeatedly proceeded until all tags in the reader field has been identified. The above description is the process of a traditional ALOHA-based tag identification method for the identification of the static tags.
When the tags are mobile, that is, the tags continually arrives at the reader field, the traditional ALOHA-based tag identification method need an adaption to resist the change of the mobile tag made. Compare with the identification of the static tags, there are two concerns about how to identify the mobile tags in this research. The first concern is that the identification of all arrival tags on the current read cycle will be delayed until the beginning of the next read cycle. This is because the reader broadcasts the frame size only at the beginning of the read cycle in the ALOHA-based static tag identification. The new arrival tags can not receive the frame size of the current read cycle immediately at the time of the tag arrives at the reader field. Therefore, the arrival tag must wait the reader broadcast the
frame size until the beginning of the next read cycle. After the tag receives the frame size, the reader then uses the frame size to choose a slot to respond the tag ID. Following the above process, a new arrival tag must wait until the reader broadcast the frame size at the beginning of the next read cycle. This phenomenon may increase tag identification time. To reduce the elapse time from the time of the tag arrival to the time of the next read cycle beginning, each slot broadcasts the slot ID and the frame size of the current read cycle at its beginning. After the new arrival tag receives the slot ID and the frame size, the tag randomly choose a slot from slot ID to slot frame size – 1. Therefore, the tag may be identified at the read cycle of it arrives at and the elapse time of the tag identification may be reduced.
The second concern is how to estimate the number of the tag to be identified on the incoming read cycle. In [4], Schoute proposed that the frame slotted ALOHA reached its optimal throughput3 when the number of slots is equal to the number of tags. The optimal throughput of slotted ALOHA is about 36.8% in the condition of the number of tags and the number of slots are unlimited. To fully use the Schoute’s view, the frame size of the read cycle has better to equals the number of unidentified tags in every read cycle. Due to the frame size of the read cycle must be decide at the beginning of the read cycle, the estimation of the number of tags which need to be identified of the read cycle is needed.
The number of the tag to be identified for the current read cycle can be estimated by using the sum of the number of the unidentified tags and the number of the arrival tags on the pervious read cycle. Therefore, when a reader complete a read cycle, the reader need to estimate the number of the unidentified tags and the number of the arrival tags of the previous read cycle to decide the frame size of the current read cycle. If the reader makes a precise estimation, the throughput of the current read cycle may be optimal.
3 Throughput = the number of identified tags/total slots used.
4.2.2. Enhanced dynamic framed slotted ALOHA for mobile tags
In this section, the enhanced dynamic framed slotted ALOHA for mobile tags is described. The structure of a frame is specified first and then the identification procedures of the reader and the tags are stated.
The frame structure of MT-EDFSA is shown in Figure 7. A period of making the identification once for every unidentified tag in a reader field is called a frame. The minimal working unit of the frame is a slot. There are two types of slots in the frame structure––unidentified tag slot and arrival tag slot. An unidentified tag slot is used to identify the unidentified tags which were in the waiting state. An arrival tag slot is used to identify the tags which are in the arriving state. Slots are grouped into segments. An unidentified tag segment contains a set of unidentified tag slots and an arrival tag segment contains a set of arrival tag slots.
Figure 7 The frame structure of MT-EDFSA
To make the mobile tag identification, a reader performs the following procedure:
1. At the beginning of a read cycle, the reader estimates the number of the identifying tags and uses it as the frame size of the current read cycle. The number of the identifying tags, F, can be represented as
F = 2.39 * N
previous, frame, collision+ 2.39* Nprevious, arrival, collision + Nprevious, arrival, success, (1) in which Nprevious, frame, collision denotes the number of the collision slots on the frame of the pervious read cycle, Nprevious, arrival, collision denotes the number of the collision slots on the arrival tag segment of the previous read cycle, Nprevious, arrival, success denotes the number ofthe success slots on the arrival tag segment of the previous read cycle. In equation 1, the first term represents the number of unidentified tags in the previous read cycle. The sum of the second term and the third term represents the number of arrival tags on the current read cycle.
2. The reader broadcasts the frame size to every tag in the reader field at the beginning of the current read cycle.
3. The reader makes the tag identification slot by slot.
3.1 If the tag is in the unidentified segment, the reader broadcasts the ID of the first and the last slot of the arrival segment on the current read cycle at the beginning of the slot. If the tag is in the arrival segment, the reader broadcasts the ID of the next slot and the next two slot at the beginning of the slot.
3.2 The reader waits tags to respond their tag ID.
(1) If two or more tags respond, no tag is identified in the slot and this slot is collision slot. The reader increases the number of the collision slots on the current read cycle. If the slot is also an arrival tag slot, the reader increases the number of the collision slots on the arrival tag segment of the current read cycle.
(2) If the only one tag responds and the slot is also an arrival tag slot, the reader increase the number of the success slots on the arrival tag segment of the current read cycle.
3.3 If the read cycle is completed, set Nprevious, frame, collision to the number of the collision slots on the current read cycle, Nprevious, arrival, collision to the number of the collision slots on the arrival tag segment of the current read cycle, Nprevious, arrival, success to the number of the success slots on the arrival tag segment of the current read cycle. The reader goes back step 1 and makes the tag identification on the next read cycle. Otherwise, if the read cycle is not completed yet, goes back step 3.1 and makes the tag identification on the next slot.
In the mobile tag identification, the tag may arrive at the reader field at any time. The frame structure of MT-EDFSA is organized as the unidentified tag segment and the arrival segment. The tag may arrive at the unidentified tag segment or the arrival segment. When the mobile tag arrives at the unidentified tag segment, the tag performs the following procedure:
1. A tag arrives at the reader field. It begins waiting until the reader broadcast the first slot ID and the last slot ID of the arrival tag segment of the current read cycle at the beginning of the next slot.
2. The tag receives the IDs of the first and the last slot IDs of the arrival segment on the current read cycle.
3. The tag randomly chooses a slot which is between the first and the last slot IDs of the arrival segment on the current read cycle for the tag ID responding.
4. The tag responds its ID at the slot which is chosen at step 3.
5. The tag wait to the reader inform the result of this identification. If the status is success, the tag is identified. The identification of the tag is done. Otherwise, if the status is collision, the tag is not identified in this slot and needs identification on the next read cycle.
6. The tag waits until the beginning of the next read cycle and receives the size of the unidentified segment of the next read cycle.
7. The tag randomly chooses a slot from the unidentified segment of the next read cycle for the tag ID responding.
8. The tag responds its ID at the slot which is chosen at step 7.
8.1 The tag wait until the reader broadcasts the range of the slot IDs the first and the last slot IDs of the arrival segment on the current read cycle.
8.2 The tag responds its ID.
8.3 The tag wait to the reader inform the result of this identification. If the status is success, the tag is identified. The identification of the tag is done. Otherwise, if the status is collision, the tag is not identified in this slot and needs identification on the next read cycle.
8.4 The tag goes to step 6 and is identified again until the tag identification is success made.
When the mobile tag arrives at the arrival tag segment, the tag performs the following procedure:
1. A tag arrives at the reader field. It begins waiting until the reader broadcast the two next and three next slot IDs at the beginning of the next slot.
2. The tag receives the two next and three next slot IDs from the arrival segment on the current read cycle.
3. The tag randomly chooses a slot which is between the two next and three next slot IDs of the arrival segment from the current read cycle for the tag ID responding.
4. The tag responds its ID at the slot which is chosen at step 3.
5. The tag wait to the reader inform the result of this identification. If the status is success, the tag is identified. The identification of the tag is done. Otherwise, if the status is collision, the tag is not identified in this slot and needs identification on the next read cycle.
6. The following steps are the same as step 7 to step 9 in the mobile tag arrives at the unidentified tag segment procedure.
4.3. Performance evaluation
In this section, the experiment was performed to evaluate the performance of MT-EDFSA through simulations. MT-EDFSA was compared with two other methods - DFSA and DFSA for mobile tags (MT-DFSA).
To evaluate the performance of these methods, a simulation model is developed in which only one reader was used to identify tags in the reader field. This research also assumed that the communication channel between the reader and tags was ideal (i.e., error-free). All tags here arrives at the reader field is following the poison process and the interval time between two tag arrival is following the exponential distribution. Tags that are identified in the read cycle will not participant in the next read cycle. Based on this model, this research performed the experiment; the performance metric in this experiment is the service time (per tag), which is the time between tag arrives at the reader field and is identified by the reader. The simulation model was implemented using JAVA on a PC running Windows XP Professional. Some parameters in the experiments were set as follows.
Total identification time in each method is set to 100000 slots.
The initial frame size is 4 slots
4.3.1. The effects of arrival rate on service time
The effects of the arrival rate on the service time among MT-EDFSA, MT-DFSA and DFSA were studied in this experiment. The arrival rate was varied from 0.1 to 0.5 by step of 0.01. The experiment results are shown in Figure 8. In Figure 8, the x-axis represents the arrival rate and the y-axis represents the service time.
0 5000 10000 15000 20000 25000 30000 35000 40000
0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
Service time
Arrival rate
DFSA MT-DFSA MT-EDFSA
Figure 8 The effects of arrival rate on service time (arrival rate: 0.1 to 0.5).
0 20 40 60 80 100 120
0.1 0.12 0.14 0.16 0.18 0.2 0.22 0.24 0.26 0.28 0.3
Service time
Arrival rate
DFSA MT-DFSA MT-EDFSA
Figure 9 The effects of arrival rate on service time (arrival rate: 0.1 to 0.3).
0 5000 10000 15000 20000 25000 30000 35000 40000
0.3 0.35 0.4 0.45 0.5
Service time
Arrival rate
DFSA MT-DFSA MT-EDFSA
Figure 10 The effects of arrival rate on service time (arrival rate: 0.3 to 0.5).
The first result was that MT-EDFSA outperformed DFSA and MT-DFSA in terms of service time. The service time of MT-EDFSA was lower than those of MT-DFSA and DFSA.
The first reason is that MT-EDFSA has a new identification procedure. Therefore, the reader is able to make the tag identification in the current read cycle. The second reason is that the number of unidentified tags in the current read cycle is precisely estimated. MT-EDFSA uses it as the frame size of the current read cycle. Following the first feature of FSA [10], MT-EDFSA has the lowest service time.
The second result that is worth our attention is that when the arrival rate becomes large, the service time will be increase rapidly. The service time of both MT-DFSA and DFSA are dramatically increase when the tag arrival rate are large than 0.31 and the service time of MT-EDFSA are dramatically increase when the tag arrival rate are larger than 0.36. The reason is that the number of arrival tags is too large and the reader can’t afford to identify them. Therefore, if the number of readers is increased, the service time of each method might be increased smoothly. The result also shows that MT-DFSA can resist more tag arrival than MT-DFSA and DFSA.
The third result is that; our proposed method - MT-EDFSA can identify more tags without add extra readers. Figure 9 and Figure 10 shows MT-EDFSA can identify tags without additional readers when the arrival rate is less or equal than 0.35. Comparing to other methods; at the same arrival rate, MT-EDFSA can identify more tags without additional readers and the service time less than others.
Chapter 5 Conclusions
This research performed two works on RFID tag identification (static tag and mobile tag). In static tag identification, we proposed an anti-collision method based on tag grouping was proposed. The proposed method was called grouping based dynamic framed
slotted ALOHA (GB-DFSA). GB-DFSA fully utilized the feature of FSA, that is, if the
number of tags is small and the frame size is equal to the number of tags, the throughput might exceed to 36.8%. This feature has not yet gotten enough attention in the research community. The strategy to fully utilize this feature in GB-DFSA is that the tags were divided into small groups. This research compared the throughput of GB-DFSA with P-EDFSA and DFSA through simulations. The results showed that the throughput of GB-DFSA is about 40%, which is higher than P-EDFSA (about 38%) and DFSA (31%).GB-DFSA outperformed P-EDFSA and GFSA in term of throughput.
In mobile tag identification, the mobile tag identification issue is explored in this research and this research presented the mobile tag identification method - enhanced dynamic framed slotted ALOHA for mobile tags (MT-EDFSA) to make the mobile tag identification. In MT-EDFSA, the reader broadcasts the ID of the first and the last slot of arrival segment of the current read cycle to arrival tag. The arrival tag, therefore, may be identified in the current read cycle and the service time of the tag identification is reduced.
Further, MT-EDFSA estimated the number of the tag to be identified on the current read cycle. MT-EDFSA set the frame size of the current read cycle to the estimated number and hope to reach the optimal throughput. This research compared the service time of MT-EDFSA with those of the other methods – MT-DFSA and DFSA. The results show that MT-EDFSA has the lowest service time among MT-EDFSA and the others two methods in different arrival rates.
References
1. K. Finkenzeller, RFID handbook, 2nd ed., John Wiley & Sons, 2003.
2. D. Shih, P. Sun, D. Yen and S. Huang, “Taxonomy and survey of RFID anti-collision protocols,” Computer Communications, Vol. 29, pp. 2150-2166, January 2006.
3. ETH Zürich, RFID Multiple Access Methods, summer semester 2004.
4. F. S. Schoute, “Dynamic frame length Aloha,” IEEE Trans. Communications Letters, Vol. 31, No. 4, pp. 565-568, April 1983.
5. J. Park, M. Chung and T. Lee, “Identification of RFID tags in Framed Slotted ALOHA with Robust Estimation and Binary Selection,” IEEE Communications Letters, Vol. 11, Issue 5, pp. 452-454, May 2007.
6. J. Park, M. Chung and T. Lee, “Identification of RFID tags in Framed Slotted ALOHA with tag estimation and binary splitting,” 1st International Conference on
Communications and Electronics, pp. 368-372, January 2006.
7. J. Cha and J. Kim, “Novel anti-collision algorithms for fast object identification in RFID system,” 11th International conference on Parallel and Distrusted Systems (ICPADS 2005), Workshops, Vol. 2, pp. 63-67, July 2005.
8. J. Wang and G. Wang, “An anti-collision algorithm using two-functioned estimation for RFID tags,” Computational science and Its Application (ICCSA 2005), Vol. 3483, pp.
702-711, May 2005.
9. S. Lee and C. Lee, “An enhanced Dynamic framed slotted ALOHA anti-collision algorithm,” International Conference on Mobile and Ubiquitous Systems: Networking
and Services (MobiQuitous 2005), pp. 166-174, 2005.
10. W. Shin and J. Kim, “Partitioning of tags for near-optimum RFID anti-collision performance,” Wireless Communications and Networking Conference (WCNC 2007), pp. 1673-1678, March 2007.
11. J. Kim, “A Divide-and-Conquer Technique for Throughput Enhancement of RFID Anti-collision Protocol,” IEEE Communications Letters, Vol. 12, No. 6, June 2008.
12. J. Eom, T. Lee, R. Rietman and A. Yener, “An efficient framed slotted ALOHA algorithm with pilot frame and binary selection for anti-collision of RFID tags,” IEEE
Communication Letters, Vol. 12, Issue 11, pp. 861-863, November 2008.
13. M. Yeh, J. Jiang and S. Huang, “Adaptive splitting and pre-signaling for RFID tag anti-collision,” Computer Communications, Vol. 32, Issue 17, pp. 1862-1870, November 2009.
14. T. Wang, “Enhanced binary search with cut-through operation for anti-collision in RFID Systems,” IEEE Communication Letters, Vol.10, No. 4, pp. 236-238, April 2006.
15. H. Lee and J. Kim, “QT-CBP: A new RFID tag anti-collision algorithm using collision bit positioning,” EUC Workshops 2006, pp.591-600, 2006.
16. J. Myung and W. Lee, “An adaptive memory-less tag anti-collision protocol for RFID networks,” IEEE INFOCOM 2005, 2005.
17. J. Myung and W. Lee, “Adaptive binary splitting: A RFID tag collision arbitration protocol for tag identification,” Mobile Networks and Applications 2006, pp.711-722, 2006.
18. J. Myung, W. Lee, J. Srivastava and T. Shih, “Tag-splitting: Adaptive collision arbitration protocols for RFID tag identification,” IEEE Transactions On Parallel And
Distributed Systems, Vol. 18, No. 6, pp. 763-775, June 2007.
19. Y. Lai and C. Lin, “A pair-resolution blocking algorithm on adaptive binary splitting for RFID tag identification,” IEEE Communications Letters, Vol. 12, No. 6, 2008.
20. Y. Lai and C. Lin, “Two blocking algorithms on adaptive binary splitting: Single and pair resolutions for RFID tag identification,” IEEE/ACM Transactions On Networking, Vol. 17, No. 3, pp. 962-975, June 2009.
21. K. Yeh, N.W. Lo, Y. Li and E. Winata, “An Adaptive n-Resolution Anti-Collision Algorithm for RFID Tag Identification,” IEEE 24th International Conference on
Advanced Information Networking and Applications Workshops, 2010.
Appendix 1 – Java code for GB-DFSA
Main function of GB-DFSA
package LinkListBase;
public class enhancedPartitionAlgorithmVer2{
public static void main(String[] args){
int tagNum=100;
int totalCount=0;
int totalGroups=0;
for(int j=0;j<1000;j++) {
int slotNum=4,subgroups=1,setIndex=1;
int collision=0,idle=0;
int tag[][]=new int[tagNum][2];
int collidAndSelect[]=new int[3];
Tag3 frame[];
int tagSelected=0,slotTotal=0;
boolean splitStatus=false;
tagPartitionAndRead tpar=new tagPartitionAndRead();
for(int i=0;i<tagNum;i++){
tag[i][0]=0;
tag[i][1]=1;
}
while(tagSelected<tagNum){
slotTotal+=slotNum;
frame=new Tag3[slotNum];
frame=tpar.slotAndGroupAllocate(tagNum, slotNum, subgroups, setIndex, tag, splitStatus);
collidAndSelect=tpar.readSlots(tag, frame, slotNum);
collision=collidAndSelect[1];
tagSelected+=collidAndSelect[0];
idle+=collidAndSelect[2];
if(collision>0){
if(setIndex==1){
int newSlot=(int)2.39*collision;
if(newSlot>=4){
subgroups*=4;
} else{
slotNum=newSlot;
} } else{
slotNum=(int)2.39*collision;
splitStatus=true;
} } else{
if(tagSelected>=setIndex) {
slotNum=(int)tagSelected/setIndex;
} setIndex++;
} }
totalCount+=slotTotal;
totalGroups+=subgroups;
System.out.println("Identified tags:"+tagSelected+"\tTotal time slot:"+slotTotal+"\tThroughput:"+(double)tagNum/slotTotal);
}
System.out.println("Average time slot:"+(int)totalCount/1000+"\tAverage throughput:"
+(double)tagNum/(int)(totalCount/1000)+"\tAverage group size:"+(int)totalGroups/1000);
} }
Sub function of GB-DFSA
package LinkListBase;
import java.util.Random;
public class tagPartitionAndRead {
public Tag3[] slotAndGroupAllocate(int tagNum,int slotNum,int subgroups,int setIndex,int tag[][],
boolean splitStatus){
Tag3 frame[];
frame=new Tag3[slotNum];
Random slotAllocate=new Random();
Random groupAllocate=new Random();
for(int i=0;i<tagNum;i++){
int slot=slotAllocate.nextInt(slotNum);
int tagGroup=0;
if(tag[i][0]==0){
if(setIndex>1){
if(splitStatus==false){
tagGroup=groupAllocate.nextInt(subgroups-1)+setIndex;
tag[i][1]=tagGroup;
} else{
tagGroup=tag[i][1];
} } else{
if(subgroups>1){
if(splitStatus==false) {
tagGroup=groupAllocate.nextInt(subgroups)+setIndex;
tagGroup=groupAllocate.nextInt(subgroups);
tag[i][1]=tagGroup;
} else {
tag[i][1]=tagGroup;
tagGroup=tag[i][1];
} } else{
tagGroup=1;
}
tag[i][1]=tagGroup;
}
Tag3 tagS=new Tag3(i,tagGroup);
if(tagS.subGroup==setIndex){
tagS.next=frame[slot];
frame[slot]=tagS;
} } }
return frame;
}
public int[] readSlots(int tag[][],Tag3 frame[],int slotNum){
int collision=0;
int tagSelected=0;
int idle=0;
int status[]=new int[3];
for(int i=0;i<slotNum;i++){
Tag3 p=frame[i];
if(p!=null){
if(p.next==null){
tag[p.tag][0]=1;
tagSelected++;
} else{
collision++;
}
} else {
idle++;
} }
status[0]=tagSelected;
status[1]=collision;
status[2]=idle;
return status;
}
Tag
package LinkListBase;
public class Tag3 { public int tag;
public int subGroup;
Tag3 next;
public Tag3(int tag,int subGroup){
this.tag=tag;
this.subGroup=subGroup;
next=null;
} }
