中華大學

(1)

中華大學

碩士論文

RFID 網路中以分層方法移除多餘讀取器之最佳化技術

A Layered Optimization Approach for Redundant Reader Elimination

in RFID Networks

系別所:資訊工程學系碩士班學號姓名: E09402013 陳逸民指導教授:許慶賢博士

中華民國九十六年七月

(2)

A Layered Optimization Approach for Redundant Reader Elimination

in RFID Networks

By

Yi-Min Chen

Advisor: Prof. Ching-Hsien Hsu

Department of Computer Science and Information Engineering Chung-Hua University

Hsin-Chu, 30067, Taiwan

July 2007

(3)

中文摘要

RFID 技術的迅速發展已促進大規模RFID系統的發展和研究，關於RFID碰撞防止和多餘的RFID 讀取器移除問題已被廣泛研究及探討。在這兩個問題領域已經有許多研究人員提出許多不同的經驗法則演算法。在這篇文章裡，我們提出分層移除之最佳化技術 (LEO)，可適用在多餘的RFID 讀取器問題上，目的是要辨識出最大數量之多餘的RFID 讀取器使其能安全的移除或關閉，且可保留原有的 RFID 網路所涵蓋的範圍。被提出的LEO是一種近似最佳化的獨立演算法技術，

LEO排程可使多餘的讀取器在辨識過程中大量降低對TAG的寫入次數。而且，在 RFID 讀取器之間不需要相互聯繫，不需要時間上的同步也不需收集整體資訊來集中控制，這都是LEO排程的優點。在演算法的效能評估上，我們製作了與其他辨識演算法混合的排程來實現LEO的技術。在我們的實驗結果方面，LEO排程演算法是有效的而且在多餘RFID 讀取器的辨識及降低演算法overhead有更好的效能。

關鍵詞：分層消除最佳化，多餘讀取器移除，RFID 網路，碰撞防止。

(4)

Abstract

The rapid advance of RFID technologies motivates the development and research of large scale RFID systems. The problem of RFID reader collision avoidance and redundant RFID reader elimination has been recently studied. Both of these two problems have instigated researchers to propose different heuristic algorithms. In this paper, we present a Layered Eliminate Optimization (LEO) technique for redundant reader problems aims to detect maximum amount of redundant readers could be safely removed or turned off with preserving original RFID network coverage. The proposed LEO scheme is an algorithm independent optimization technique. A significant improvement of the LEO scheme is that number of write operations could be largely reduced in the redundant reader identification phase. Moreover, there is no need of communications between RFID readers, no need of time synchronization and no need of collecting global information for centralize control are also advantages of this approach. To evaluate the performance of the proposed algorithms, we have implemented the LEO technique along with other redundant reader identification algorithm and other composite schemes. In our experimental results, the LEO is shown to be effective and provides superior performance in terms of larger number of redundant reader detected and lower algorithm overheads.

Keywords: layered eliminate optimization, redundant reader elimination, RFID network, reader collision avoidance

(5)

Acknowledgements

First of all, I would like to thank Ching-Hsien Hsu professor and every oral examination professor, give me this chance to participate in the oral examination of graduation.

Prof. Ching-Hsien Hsu is one’s adviser. He is a conscientious and careful scholar.

He also gives lots of suggestions not only for the thesis but also for my life of graduate.

One is fortunate to be one of Prof. Hsu’s graduate student. I would also like to thank members of P.D. Lab, they always give me support when one works on the thesis.

Finally, I would like to thank my family to give me great support, show tolerance for me and give me power to get through these 2 years.

(6)

List of Figures

Figure 1: An example of wireless RFID network with redundant reader ...9

Figure 2: Interaction flow of redundant reader identification in RRE algorithm ...12

Figure 3: Interaction flow of redundant reader identification in LEO algorithm ...14

Figure 4: Second example of wireless RFID network with redundant reader...15

Figure 5: Redundant reader identification for the second example (a) result of RRE (b) result of LEO...16

Figure 6: Interaction flow of redundant reader identification in LEO+RRE algorithm18 Figure 7: Third example of wireless RFID network with redundant reader ...19

Figure 8: Redundant reader identification for the third example (a) result of RRE (b) result of LEO ...20

Figure 9: Resulting network topology of the third example (a) after RRE is run (b) after LEO is run ...20

Figure 10: Redundant reader identification for the third example (a) result of RRE+LEO (b) result of LEO+RRE ...21

Figure 11: Snapshot of RFID network simulator (a) a randomly generated RFID network (b) network topology after redundant readers are removed ...22

Figrue 12: Comparison of redundant reader detected with network area 1000010000, reader radius=500 and number of reader=500. ...24

Figure 13: Comparison of number of write operations with network area 1000010000, reader radius=500 and number of reader=500. ...25

Figure 14: Comparison of redundant reader detected with network area 1000010000, number of tags=4000 and number of reader=500 ...26

(8)

List of Figures

Figure 15: Comparison of number of write operations with network area

1000010000, number of tags=4000 and number of reader=500 ...27 Figure 16: Comparison of number of redundant reader detected (a) and accumulated

(b) with network area 1000010000, reader radius=500, number of

tags=4000, reader=500...27 Figure 17: Fourth example of wireless RFID network with redundant reade ...28 Figure 18: Comparison of redundant reader detected with network area 1000010000,

reader radius=500 and number of reader=500. (RFID tags randomly

deployed increases from 500 to 3000) ...31 Figure 19: Comparison of redundant reader detected with network area 1000010000,

reader radius=500 and number of tags=4000. (RFID readers randomly deployed increases from 500 to 4500) ...31

(9)

CHAPTER 1 Introduction

1.1 Motivation

Radio Frequency Identifier (RFID) System is an automatic technology aids machines or computers to identify, record or control individual target through radio waves. An RFID system is composed by two components, tags and readers. An RFID tag is comprised of integrated circuit with an antenna for storing information and communication, respectively. An RFID reader is capable of reading the information stored at tags located in its sensing range. The electronics in the RFID reader use an outside power resource to generate signal to drives the reader’s antenna and turn into radio wave. The radio wave will be received by RFID tag which will reflect the energy in the way of signaling its identification and other related information. In some matured systems, the reader’s RF can also instruct the memory to be read or write from which the tag contained.

Many applications, such as supply chain automation, identification of products at check-out points, security and access control, have been developed to take the primary function of RFID systems. Advantages of RFID technologies, such as price efficiency and accuracy of stock management also broaden the scope of applications of RFID systems. Furthermore, some advanced characteristics of recent RFID readers, like size miniaturization and capabilities of Wi-Fi or cellular also motivate the development of large scale RFID systems.

It will be necessary to install RFID readers in appropriated distance to each other in many applications. Otherwise it would be interfere with each other. The interference

(10)

could be caused when the frequency band is shared with other potential users. As an RFID reader is designed to accept the tiny signal reflected from a tag. It will be

particularly influenced to any relatively powerful transmissions from other readers that happen at the same time.

Similar to wireless sensor network, RFID tags can be deployed in an ad-hoc fashion instead of pre-installed statically. It is motivated in recent RFID technologies that an RFID system can be integrated with wireless sensor network by interfacing RFID tags with external sensing capabilities, such as light, temperature or shock sensors.

In such way, the hybrid infrastructure combines advantages of both techniques, such as accurate identification, monitoring of objects and efficient deployment.

1.2 Objective

This study addresses the problem of redundant reader elimination. The problem of redundant reader elimination was proved as NP-hard problem [2, 3]. In this thesis, we propose a Layered Eliminate Optimization (LEO) heuristic to identify redundant readers. Objective of this optimization is to detect the maximum number of redundant readers that can be safely turned off with preserving the origin network coverage in an RFID network.

To evaluate performance of the proposed technique, we have implemented a simulator with random graph generator. The simulation results demonstrate that LEO provides superior performance in terms of number of redundant reader detected. Both theoretical analysis and experimental results show that LEO has lower algorithm overheads, i.e., number of write operations issued by readers. In short, LEO is suitable in arbitrary RFID network and applicable to real RFID environment in practice.

(11)

1.3 Thesis Organization

The rest of this thesis is organized as follows: In Chapter 2, a brief survey of related work will be presented. Chapter 3 introduces the redundant reader problem.

The layered eliminate optimization for redundant reader minimization will be discussed in Chapter 4. Performance analysis and simulation comparisons will be given in Chapter 5. Finally, in Chapter 6, some concluding remarks are made.

(12)

CHAPTER 2 Related Work

RFID technologies can be classified into several aspects. First, security and privacy related literatures [9, 11, 17] focused on methods of preserving and protecting privacy of RFID tags; the reader collision avoidance and hidden terminal problems were first addressed in [7] aims to enhance accuracy of RFID systems; the energy saving and coverage problem were extensively studied [4, 12, 14, 19, 20] in order to improve lifetime of wireless topology network. Because this thesis is related to reader collision and coverage problem, we will not describe details of security and privacy issues in this chapter.

Research efforts for collision avoidance have been presented in the literature.

Frequency Division Multiple Access (FDMA), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA) and Carrier Sense Multiple Access (CSMA) are four basic access methods to categorize MAC-layer protocols.

FDMA is functioned via frequency assignment in which the communication is applied in form of many-to-one. Since RFID tags without a frequency tuning circuitry, reader selection is not allowed during communication. Therefore, the additional of such a tuning circuitry will increase the cost of the RFID tags and deployment of RFID systems. CDMA uses spectrum modulation techniques that based on pseudo random codes for data transmission. It is more complicated and computation intensive due to additional circuitry to the tags which also bring up the cost of tags. TDMA uses time slot to avoid collision and clock synchronization complement each other. It is useful in ensuring that messages sent for clock synchronization do not collide. And, clock

(13)

synchronization helps to reduce the clock drift and to ensure that the clock drift does not cause TDMA slots of nearby sensors to overlap. Because there is only one code transmitted during each slot, it allows readers to communicate using different time slot that successfully avoid the collision. To accommodate better read rate in dynamic RFID system, time slot should be reshuffled adaptively. CSMA enables individual data transmission by detecting whether the medium is busy. For hidden terminal problems in RFID networks, the interference at the tags’ communication remains due to readers may not in other’s sensing range. Thus, CSMA is not able to avoid collisions caused by hidden terminal in RFID networks.

Standard collision avoidance protocols like RTS-CTS [13] cannot be directly applied to RFID systems due to the reason, in traditional wireless networks, the CTS are sent back to the sender. Similar situation in RFID system, when a reader broadcasts an RTS, all tags in the read range need to send back CTS to the reader. It then requires another collision avoidance mechanism for CTS, and it will make the protocol more complicated.

Techniques for resolving RFID reader collision problems are usually proposed as reader anti-collision techniques or tag anti-collision solutions. The Colorwave [15] is a scheduling-based approach prevents RFID readers from simultaneously transmitting signal to a RFID tag. The Colorwave is used as a distributed anti-collision system based on TDMA in RFID network. In Colorwave, RFID network is modeled as an undirected graph. Vertices are used to represent readers, and the edges represent collision constraints. When two readers connected with an edge and transmitting data at the same time, collision occurred. The Colorwave is to address RFID network with graph coloring theory and accomplishing each reader has the smallest possible number of adjacent nodes in the same color. This approach makes easy reservation of time slots for collision-free transmission of data (i.e. reader-tag communication). A reader

(14)

will transmit only in its color (timeslot) and if the transmission collides with another reader, the transmission request is discarded. The Colorwave protocol attempts to optimize the graph to achieve a percentage of successful transmissions. However, there are still some limitations. Such as colors are selected randomly, there is no information given on how the network built up. The given feedback from the percentage of successful transmissions is used to regulate the parameters. However this will not ensure the system stays in a stable condition. The maximum number of colors in the network will not be the same in all readers. As the ratio of collisions of all regional readers exceeds a safe threshold, readers will cause the “kick wave” and then the value of maximum colors could be changed. In mobile environment, the reader stays in and out in which the behavior causes the whole topology a dynamic change.

This leads the reformation of time slot and therefore decrease system throughput.

Furthermore, readers in communicating state of time slot require synchronization. The collision might be also happed because of limited broadcast range makes the reader incapable of knowing other existing readers, i.e., hidden terminal problems.

Pulse protocol [1] is referred as beacon broadcast and CSMA mechanism [16].

Readers periodically in separated control channels send a “beacon” during communication with tags. The contend_back-off and the delay_before_beaconing in the protocol are similar in wireless networks. If the reader receives a beacon, the residual back-off timer will be stored and kept till the next coming chance. This process is expected to achieve the fairness among all readers.

Communications among readers in the Pulse protocol is achieved through a higher power in the control channel than in the data channel. The control channel is a sub-band in the RFID spectrum that is separated from reader-tag communication. The data channel is for reader-tag communication while the control channel is for reader-reader communication. The transmission in the control channel can not affect

(15)

the communication in the data channel. It is assumed that the reader is able to receive on both the control and the data channel simultaneously. In [1], the Beacon Range Factor (BRF) is defined as the ratio of the control channel transmission power to the data channel transmission power. The Pulse protocol was examined and shown effectiveness by adjusting the control channel power to cover reader range in solving Reader Collision Problems.

A coverage-based RFID reader anti-collision mechanism was proposed in [10].

Kim et al. presented a localized clustering coverage protocol for solving reader collision problems occurring among homogeneous RFID readers.

HiQ [8], an online learning algorithm, is used to find dynamic solutions to the Reader Collision Problem in RFID systems. The focus of the HiQ algorithm contains two parts: first, HiQ is used to allocate resources to maximize the number of readers communicating at a single time period; secondary, HiQ is used to minimize the number of collisions these readers experience when communicating.

HiQ utilizes three basic hierarchical tiers in its control structure: readers, R-servers, and Q-servers. The lowest tier is the RFID readers which they communicate solely when they have been granted a frequency and time slot in communication by a server (R-server) tier. R-servers are allocated frequencies and time slots by the Q-learning servers, or Q-servers. Q-learning servers comprise the highest tier in this hierarchical algorithm. Q-servers distribute resources to the servers directly below them in the hierarchy. Regardless of how many Q-server tiers there are, there is always a single root server that has global knowledge of all frequency and time slot resources, and is able to allocate all of them. This approach has some limitations: first, the protocol reserves a hierarchical structure that requires an extra management; second, to mobile readers, the topology may change the hierarchical structure which will require the distribution of time slots to be reshuffled. That will consume time and make the

(16)

system unavailable; furthermore, Q-learning assumes collision detection of readers is not in sensing range. It is believed that some collisions can not be detected and lead to incorrect operation of the protocol; finally, the use of timeslots requires all readers to be synchronized in the whole system is also the weakness of this approach.

From what we have learned in the literature about developed protocols, which are called medium access protocols, are defined to coordinate the use of a shared medium with certain regulation. However, the existing method under the mobility environments has several limitations.

The problem of coverage problem has been variety studied. Tian et al. [14]

proposed techniques for detecting redundant sensors whose coverage area is overlapped with others. In [18, 19], Ye et al. presented an energy conserving protocol to extend lifetime of wireless sensor network. The concept of working set was applied in their approach to alternatively turn sensors off and on. A centralized algorithm was proposed in [12] for organizing sensor network in disjoint subsets of sensors, in order to maximize efficient use of batteries. On the contrary, Zhang et al. [20] proposed a grid based distributed algorithm for maintaining coverage and connectivity. Preserving network coverage and eliminating redundancy in a sensor network, energy efficiency could be improved. Carbunar et al. [2] proposed an approximation algorithm for extending lifetime of wireless RFID reader network. Probabilistic analysis and experimental results report that the RRE heuristic is effective in arbitrary topology.

(17)

CHAPTER 3 Preliminary

3.1 Redundant Reader Problem

A reader is said redundant if all tags in its covered area are also covered by at least one of the other readers. Figure 1 shows an RFID network contains three readers, R1-R3, and five tags, T1-T5. Reader R2 is referred as a redundant reader because the three tags it covered, i.e., T1, T2 and T3, are also covered by other readers in the same network. Therefore, reader R2 can be safely removed without loss of tags been covered. Advantages of removing redundant readers are twofold; First, because of the limited battery associated with readers, it can extend the lifetime of overall wireless RFID network if the redundant readers are turn off alternatively; Second, the reader to reader interference could be alleviated by eliminating redundant readers.

Consequently, reader collisions could be dispelled while the monitoring accuracy of RFID network can be also improved.

T1 T2

T3

T4

T5

R1 R2 R3

Figure 1: An example of wireless RFID network with redundant reader

(18)

Objective of the redundant reader problem is to detect maximum redundancy of RFID readers in wireless RFID network. A naive method is to have all readers broadcast a query message to all its covered tags simultaneously. Because RFID tags will reply queries by signaling its id, therefore, if a reader receives no reply, it means that itself a redundant reader. This is either because the reader covers no tag in its covered range, or because tags are not able to reply due to reader collisions.

There are drawbacks of the above method to detect reader redundancy. Firstly, time synchronization among readers is required. Second, network coverage may be destroyed and resulting additional tags uncovered if all redundant readers are turned off.

The second situation can be explained by taking the same network topology shown in Figure 1. We assume the same readers, R1-R3 and only four tags T1-T4 in the RFID network. According to the above description, readers R2 and R3 will receive no tag reply and treat itself a redundant reader. Therefore, if readers R2 and R3 are both turned off, it will result tags T3 and T4 uncovered.

(19)

3.2 Research Architecture

Before we start to discuss the proposed layered eliminate optimization (LEO), we first clarify our network model, research architecture and briefly describe characteristics of LEO.

 LEO is applicable to arbitrary RFID network topology with unlimited number of RFID readers and tags.

 LEO is distributed. There is no need to collect network information for centralized control and no need to do time synchronization. Each reader performs redundancy check locally.

 No communication among RFID readers is required in LEO.

 Tags are passive and writable

 Reader collision problem is assumed avoided before to run LEO.

(20)

CHAPTER 4 Layered Eliminate Optimization

As mentioned in the subchapter of research architecture, LEO has advantages in practice, such as it is designed for arbitrary RFID network topology; there is no communication between RFID readers; and there is no need to perform time synchronization. The only assumption in LEO implementation is that reader collisions are avoided before executing LEO. Since we have previous published papers discussing contention-free transmissions of RFID readers to avoid collisions that caused by hidden terminal. We will not discuss the phase of collision avoidance in this chapter.

To verify beneficial of the proposed LEO technique, we briefly review the redundant reader elimination (RRE) method which was previously presented in [2]. To simplify the presentation, we depict the operation of redundant reader identification in the RRE algorithm as an interaction flow shown in Figure 2.

Figure 2: Interaction flow of redundant reader identification in RRE algorithm

The concept of RRE algorithm is to record “tag count”, number of tags a reader covers, onto the tag. Only the reader has maximum tag count could be holder of the

(21)

corresponding tag. Therefore, at the beginning, an RFID reader (Ri) will send a query for accumulating number of tags in its vicinity. Then, it queries all of these tags for their holder, the reader who wrote maximum tag count onto the tag. If tag count of the holder is smaller than Ri’s tag count, Ri writes its tag count onto the tag and records its id to tag’s holder. Consequently, reader Rj will be regarded as redundant if holders of all its covered tags are not Rj. In such mechanism, each tag will be written at least two data. That means “tag count” and “holder” of a tag might be updated by a later query RFID reader if the reader’s tag count is larger than previous value.

The layered eliminate optimization simplifies the above method. An RFID reader only writes reader id onto a tag. Moreover, once a tag is written by other reader, the later query RFID reader will not overwrite the tag. Therefore, the total number of write operation will at most equal to the number of RFID tags in the RFID network.

Figure 3 shows the operations of redundant reader identification in LEO algorithm.

The term “layered” represents the relationship between early query RFID readers with the late query ones. Relatively, for later query reader, it will have higher probability to be redundant.

Referring to the operations of LEO outlined in Figure 3. An RFID reader (Ri) broadcasts a query message to tags in its vicinity asking its holder. As a tag replies its holder, there are two possibilities, holder=”NULL” or holder=”Rk”, Rk is one of the readers in the RFID network. If holder=”NULL”, Ri writes its id onto the tag’s holder.

If holder=”Rk” and RkRi, Ri skips the reply. Therefore, an RFID reader will be regarded as redundant if it receives all non-NULL tag replies.

(22)

Figure 3: Interaction flow of redundant reader identification in LEO algorithm

The algorithm of LEO is given as follows.

‘r = number of readers, t = number of tags 1. For r = 1 To Num_reader

2. For t = 1 To Num_tag

3. ‘Calculate the coordinate(X,Y) of TAG(t) 4. X = Abs(Tag_(t, CX) - Reader(r, CX)) 5. Y = Abs(Tag_(t, CY) - Reader(r, CY)) 6. tag_radius = Sqr(X ^ 2 + Y ^ 2)

7. ‘ Whether confirm tag in the range of reader 8. If tag_radius < reader_radius Then

9. If Tag_(t, rid) = empty Then 10. ‘ Write reader id onto tag (t) 11. Tag_(t, rid) = r

12. ‘ Set redundant flag to false 13. Reader(r, redundant) = false 14. End If

15. End If 16. Next 17. Next

The algorithm of LEO is implemented by Visual Basic language. In beginning of the algorithm, check the tag is in its vicinity or not by each reader (lines 3-6), if t tag is in the region of reader r, the reader r query the tag’s rid, if rid of tag is empty the reader will write its id onto tag’s rid and set the redundant flag to false (lines7-13)

(23)

Let’s demonstrate the identification of redundant reader in both RRE and LEO by the example as shown in Figure 4. Figure 5(a) illustrates tag contents modified by each RFID reader while issuing a query/write operation. According to the scenario given in Figure 5(a), reader R1 firstly writes its tag count (TC) = 2 and id into tags T1

and T2 (first row of the table). Then, R2 writes (TC, Rid) = (3, R2) into all its covered tags, T3, T4 and T5 as shown in the second row. Following, reader R3 attempts to write (TC and Rid) into its covered tags, T2, T3, and T4. Because tags T3 and T4 have the same tag count which was written by reader R2 as compared to R3’s tag count, and T2’s tag count of its holder is smaller than R3’s tag count, therefore, R3 will overwrite its (TC, Rid) in T2. Finally, these five tags will be locked by readers, R1, R3, R2, R2, R2, respectively. That means no redundant reader could be detected.

Figure 4: Second example of wireless RFID network with redundant reader

Consider again the example and applying same scenario by using LEO redundant reader identification scheme. According to the interaction flow described in Figure 3, reader R1 firstly marks tags T1 and T2 as its responsible tags. Then, reader R2 writes its id as holder of the three tags in its vicinity. Following operations of R1 and R2, reader R3 will not issue write operation to tags T2, T3 and T4 because of their non-null holder.

Consequently, tags are finally held by reader R1 and R2 which results reader R3 is

(24)

redundant as shown in Figure 5(b). This example shows that LEO detects one redundant reader which was not detected by the RRE algorithm. It is worthy to mention that different order of queries by RFID readers might have different results.

1

)

R

2

(3, R

reader R

3

(b)

Figure 5: Redundant reader identification for the second example (a) result of RRE (b) result of LEO

In practice, LEO is an algorithm independent optimization technique. It can be executed either independently (one phase scheme) or combined with other redundant reader elimination methods (i.e., two phases scheme) to enhance algorithm performance.

Let’s consider another example shown in Figure 7 to demonstrate this feature.

(25)

Applying the RRE and LEO algorithms to perform redundant reader checking, Figures 8(a) and 8(b) show the results after execution of RRE and LEO algorithms, respectively.

Both of the two algorithms detected one redundant reader. By removing the redundant reader, the resulting RFID network topologies are given in Figures 9(a) and 9(b), respectively. If we apply the two algorithms to the reduced RFID networks obtained in Figure 9 interchangeably, i.e., execute LEO in Figure 9(a) and execute RRE in Figure 9(b), the results of second round redundant reader identification performed by RRE+LEO and LEO+RRE schemes are shown in Figures 9(a) and 9(b), respectively.

We observe that execution of two phases redundant reader identification in both cases detected one more redundant reader as compare to the single phase algorithm, either RRE or LEO. Due to this reason, two phases scheme is expected has superior performance in term of total number of redundant reader detected. However, two phases scheme might have higher algorithm overheads because it writes tag count and reader id in both phases. As mentioned earlier, LEO has lower algorithm overheads (i.e, number of write operations) than RRE. For the two composite approaches, LEO+RRE and RRE+LEO, we will recommend to use LEO+RRE in practice. Reason for this is because most of redundant readers can be removed by LEO if it is executed before RRE. In such way, there will have less number of readers in RFID network and minimize overheads of the RRE algorithm in the second phase. We shown the Interaction flow of LEO+RRE in Figure 6, LEO +RRE scheme has been implemented based on RRE algorithm. In the first phase, holder of some tags will be written into by LEO processes. After finishing LEO procedure, some RFID readers will be turn off, which are not hold at least one tag. In second phase, only remaining readers could execute RRE procedure, in first procedure of RRE in second phase, only the reader has maximum tag count could be writes the tag count or id onto the corresponding tag.

Before writing id onto tag, an RFID reader (Ri) compare the holder of tag, if holder =

(26)

“NULL” or RiRk , Ri writes its id onto corresponding tag. If Ri=Rk , skip the id write operation. Therefore, an RFID reader will be regarded as redundant if reader that has held no tag can be turn off without loss any tag. A detailed analysis will be discussed in next chapter regarding algorithm overheads comparison.

Figure 6: Interaction flow of redundant reader identification in LEO+RRE algorithm

The algorithm of LEO+RRE is given as follows.

' ======= Phase 1 : LEO sub-routine ===========

1. For r = 1 To Num_reader 2. For t = 1 To Num_tag

7. ‘ Whether confirm tag in the range of reader 8. If tag_radius < read_radius Then

9. ‘ if rid in tag is empty then update rid by reader(r) 10. If Tag_(t, rid) = empty Then

11. ‘ Write reader id onto tag (t) 12. Tag_(t, rid) = r

13. ‘ Set redundant flag to false 14. Reader(r, redundant) = false 15. End If

16. ‘Update Tag count for RRE of second phase 17. Reader(r, TC) = Reader(r, TC) + 1

18. End If 19. Next 20. Next

Reader Query

Reply its existence -Accumulate Tag

count, TCR

-If holder=NULL Write (Rid) to holder Accumulate

holder

If All holder≠Rid Turn off reader

Query Reply

{Tag count(TCt), holder}

-If TCR>TCt

Write(Rid) into (holder) Write(TCR) into (TCt) -If holder≠Rid

Accumulate holder If All holder≠Rid

Turn off reader

Tag

(27)

'======= Phase2 : RRE sub-routine ===========

1. For r = 1 To Num_reader

2. ‘ Only remain reader could update tag count and rid 3. If Reader(r, redundant) = false Then

4. For t = 1 To Num_tag

9. ‘ Whether confirm tag in the range of reader 10. If tag_radius < read_radius Then

11. If Tag_(t, TagCounter) < Reader(r, TagCounter) Then 12. ‘ Write Tag count onto tag(t)

13. Tag_(t, TagCounter) = Reader(r, TagCounter) 14. ‘Check the rid of tag (t) is the same as reader or not.

15. If Reader(r, rid) <> Tag_(t, rid) Then 16. ‘ Update rid in tag(t)

17. Tag_(t, rid) = r 18. End If

19. End If 20. End If 21. Next 22. End If 23. Next

Figure 7: Third example of wireless RFID network with redundant reader

(28)

T1 T2 T3 T4

R1 (2, R1) (2, R1)

R2 (1, R2)

R3 (2, R3)

R4 (2, R4)

Final (2, R1) (2, R1) (2, R3) (2, R4) Redundant

reader R2

(a)

T1 T2 T3 T4

R1 R1 R1

R2 R2

R3

R4 R4

Final R1 R1 R2 R4

Redundant

reader R3

(b)

Figure 8: Redundant reader identification for the third example (a) result of RRE (b) result of LEO

(a) (b)

Figure 9: Resulting network topology of the third example (a) after RRE is run (b) after LEO is run

(29)

T1 T2 T3 T4

R1 R1 R1

R4 R4 R4

R3

R2 -

Final R1 R1 R4 R4

Redundant

reader R2, R3

(a)

T1 T2 T3 T4

R1 (2, R1) (2, R1)

R4 (2, R4) (2, R4)

R2

R3 -

Final (2, R1) (2, R1) (2, R4) (2, R4) Redundant

reader R2, R3

(b)

Figure 10: Redundant reader identification for the third example (a) result of RRE+LEO (b) result of LEO+RRE

(30)

CHAPTER 5 Performance Evaluation and Results

In this chapter, we first introduce the simulator, a random RFID network generator and explain metrics for performance comparison. Then, we will show the simulation results.

5.1 Simulator and Comparison Metrics

To evaluate performance of the proposed optimization technique, we have implemented a random RFID network generator to simulate circumstances with different characteristics. The simulator uses the following parameters to produce arbitrary network topology. The proposed LEO method was compared along with the RRE and the composite algorithm, LEO+RRE.

(a) (b)

Figure 11: Snapshot of RFID network simulator (a) a randomly generated RFID network (b) network topology after redundant readers are removed

(31)

Figure 11 shows snapshots of the simulator used in this thesis. The simulator uses three input parameters, number of readers, number of tags, and reader radius.

Three optimization approaches, RRE, LEO, LEO+RRE, were tested in the experiments.

The simulation results will report number of redundant reader detected and number of write operation issued in each algorithm. Figure 11(a) shows snapshot of a randomly produced RFID network with 500 readers before redundancy checking. Figure 11(b) shows the snapshot of the network after redundant readers are removed.

As mentioned earlier, the objective of redundant reader problem to detect maximum number of readers that can be removed safely without changing network coverage.

Therefore, we will evaluate the number of redundant readers detected by each algorithm.

In addition, we also evaluate algorithm complexities by calculating number of write operations issued by RFID readers. In short, the larger number of redundant reader detected and lower number of “write” operation issued by readers, the algorithm is better.

5.2 Experiment Results

The first experiment was performed with constant radius of readers and variant number of RFID tags. The number of redundant readers can be detected in each algorithm is reported. Figure 12 shows LEO outperforms RRE in terms of the redundant reader detected. However, both techniques detect less number of redundant readers than the composite approach. As mentioned in last chapter, RRE+LEO scheme is not recommended in practice due to its very high algorithm overheads. Therefore, our experiments only show the results of LEO+RRE.

(32)

Figure 12: Comparison of redundant reader detected with network area 1000010000, reader radius=500 and number of reader=500.

Figure 13 shows comparison of algorithm overheads in different schemes. Algorithm overhead is referred as number of write operation issued by all readers in RFID network in order to perform redundant reader identification. We observe that algorithm overheads has the order LEO < LEO+RRE < RRE. This phenomenon matches our expectation. LEO has least number of write operations and RRE performs worst. A simple reason for this result is because LEO has (m) as upper bound of write operation while RRE has (2m) as lower bound if there are m RFID tags in the network.

Furthermore, if an RFID tag is covered by r readers on average, RRE will have (2mr) as lower bound of write operation while LEO remains (m) as upper bound. Reason for LEO+RRE has less algorithm overheads than RRE is because LEO uses the same memory space Rid and removes most of redundant readers in the first algorithm phase, the overheads of RRE algorithm could be largely reduced in the second phase. The LEO+RRE will have (2m(r-d)) as lower bound of write operation, if there are d RFID readers removed by LEO.

110 130 150 170 190 210 230 250 270 290 310

1000 2000 3000 4000 5000 6000 7000 8000 9000

CHAPTER 6 Conclusions and Future work

In this thesis, we have presented a distributed optimization technique, LEO for addressing the redundant reader problem in wireless RFID network. LEO is an algorithm independent optimization technique which is applicable in arbitrary RFID network topology. To evaluate performance of the proposed technique, we have compared the LEO method along with other redundant reader identification algorithm as well as other composite schemes. The experimental results show that LEO provides superior performance in terms of larger number of redundant reader detected and lower number of write operations, i.e., algorithm overheads. The LEO scheme is verified effective under high density wireless RFID reader network.

In the future, energy saves to be a kind of trend, as we know, the main power consumption is from the reader operation. Therefore, to minimize the reader operation is necessary, in our future research, beside subject of redundant reader problem, how to reduce reader collision and write operation of reader are also important.

(39)

Appendixes

100 125 150 175 200 225 250 275 300 325 350 375 400

500 1000 1500 2000 2500 3000

Reference

[1] Shailesh M. Birari and Sridhar Iyer, “Mitigating The Reader Collision Problem in RFID Networks”, International Conference on Networking (ICON), 2005.

[2] Bogdan C˘arbunar, Murali Krishna Ramanathan, Mehmet Koyut¨urk, Christoph Hoffmann and Ananth Grama, “Redundant Reader Elimination in RFID Systems,”

Proceedings of IEEE SECON, pp. 176-184, 2005.

[3] Bogdan C˘arbunar, “Coverage problems in wireless sensor and RFID system,”

Technical Report 2005-40 Purdue University 2005.

[4] M. Cardei and J. Wu. “Coverage in Wireless Sensor Networks. Handbook of Sensor Networks,” CRC Press, 2004

[5] J. R. Cha and J. H. Kim, “Novel Anti-collision Algorithms for Fast Object Identification in RFID system,” in Proc. ICPADS2005, Fukuoka, Japan, Jul. 20-22, 2005, pp. 63-67

[6] J. R. Cha and J. H. Kim, “Dynamic Framed Slotted ALOHA Algorithm using Fast Tag Estimation method for RFID System,” in Proc. CCNC2006, Las Vegas, USA, Jan. 8-10, 2006.

[7] D. W. Engels and S. E. Sarma, “The reader collision problem,” Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, 2002.

[8] Junius Ho, Daniel W. Engels and Sanjay E. Sarma, “HiQ: A Hierarchical Q-Learning Algorithm to Solve the Reader Collision Problem,” Proceedings of the International Symposium on Applications and the Internet Workshops (SAINTW'06), pp. 88-91, 2006.

[9] Ari Juels, Ronald L. Rivest, and Michael Szydlo, “The blocker tag: selective blocking of RFID tags for consumer privacy,” Proceedings of the 10th ACM Conference on Computer and communications security, pp. 103–111, 2003.

[10] Joongheon Kim, Sunhyoung Kim, Dongshin Kim, Wonjun Lee and Eunkyo Kim,

“Low-Energy Localized Clustering: An Adaptive Cluster Radius Configuration Scheme for Topology Control in Wireless Sensor Networks,” Proceedings of the IEEE Vehicular Technology Conference (VTC), pp. 2546-2550, 2005.

(41)

[11] Sanjay E. Sarma, Stephen A. Weis, and Daniel W. Engels, “RFID systems and security and privacy implications,” proceedings of the 4^th international workshop on Cryptographic Hardware and Embedded Systems, LNCS 2523, pp. 454–469.

[12] Sasa Slijepcevic and Miodrag Potkonjak, “Power efficient organization of wireless sensor networks,” Proceedings of IEEE ICC, pp. 472-476, 2001.

[13] Joao Luıs Sobrinho, Roland de Haan and Jose Manuel Brazio, Why RTS-CTS is not your ideal wireless LAN multiple access protocol,” Proceedings of the IEEE Wireless Communications and Networking Conference, pp. 81-87, 2005.

[14] Di Tian and Nicolas D. Georganas, “A coverage-preserving node scheduling scheme for large wireless sensor networks,” Proceedings of the 1st ACM WSNA’02, pp. 32–41, 2002.

[15] James Waldrop, Daniel W. Engels and Sanjay E. Sarma, “Colorwave: an anticollision algorithm for the reader collision problem,” Proceedings of the IEEE International Conference on Communications, pp. 1206-1210, 2003.

[16] Xin Wang and Koushik Kar, “Throughput modelling and fairness issues in CSMA/CA based ad-hoc networks,” Proceedings of the 24th Annual Joint Conference on IEEE Computer and Communications Societies (INFOCOM 2005), pp. 23 – 34, 2005.

[17] Stephen Weis, Sanjay E. Sarma, Ronald L. Rivest and Daniel W. Engels, ”Security and privacy aspects of low-cost radio frequency identification systems,” Security in Pervasive Computing, LNCS 2803, pp. 201-212, 2004.

[18] Fun Ye, Shiann-Tsong Sheu, Tobias Chenand Jenhui Chen, “The Impact of RTS Threshold on IEEE 802.11 MAC Protocol,” Tamkang Journal of Science and Engineering, Vol. 6, No. 1, pp. 57-63 2003.

[19] Fan Ye, Gary Zhong, Songwu Lu and Lixia Zhang, “Peas: a robust energy conserving protocol for long-lived sensor networks,” Proceedings of the 23^rd IEEE ICDCS, pp.28-37, 2003.

[20] Honghai Zhang and Jennifer C. Hou, “Maintaining sensing coverage and connectivity in large sensor networks,” Wireless Ad Hoc and Sensor Networks: An International Journal, Vol. 1, No. 1-2, pp. 89-123, January 2005.

中 華 大 學

中 華 大 學

碩 士 論 文

RFID 網路中以分層方法 移除多餘讀取器之最佳化技術

A Layered Optimization Approach for Redundant Reader Elimination

in RFID Networks

系 別 所:資訊工程學系碩士班 學號姓名: E09402013 陳逸民 指導教授:許 慶 賢 博士

中華民國 九十六 年 七 月

A Layered Optimization Approach for Redundant Reader Elimination

in RFID Networks

By

Yi-Min Chen

Advisor: Prof. Ching-Hsien Hsu

Department of Computer Science and Information Engineering Chung-Hua University

Hsin-Chu, 30067, Taiwan

July 2007

中文摘要

Abstract

Acknowledgements

Table of Contents

List of Figures

List of Figures

CHAPTER 1 Introduction

1.1 Motivation

1.2 Objective

1.3 Thesis Organization

CHAPTER 2 Related Work

CHAPTER 3 Preliminary

3.1 Redundant Reader Problem

3.2 Research Architecture

CHAPTER 4

Layered Eliminate Optimization

T

T

T

T

T

R

(2, R

) (2, R

)

R

(3, R

) (3, R

) (3, R

)

R

(3, R

)

Final (2, R

) (3, R

) (3, R

) (3, R

) (3, R

) Redundant

reader none

T

T

T

T

T

R

R

R

R

R

R

R

R

Final R

R

R

R

R

Redundant

reader R

CHAPTER 5

Performance Evaluation and Results

5.1 Simulator and Comparison Metrics

5.2 Experiment Results

中華大學

中華大學

碩士論文

RFID 網路中以分層方法移除多餘讀取器之最佳化技術

系別所:資訊工程學系碩士班學號姓名: E09402013 陳逸民指導教授:許慶賢博士

中華民國九十六年七月