04114235

download 04114235

of 3

Transcript of 04114235

  • 8/12/2019 04114235

    1/3

    IEEE COMMUNICATIONS LETTERS, VOL. 11, NO. 1, JANUARY 2007 85

    Query Tree-Based Reservation for Efficient RFID Tag Anti-Collision

    Ji Hwan Choi, Student Member, IEEE, Dongwook Lee, and Hyuckjae Lee, Member, IEEE

    Abstract Since the tree based RFID tag anti-collision protocol

    achieves 100% read rate, and the slotted ALOHA based taganti-collision protocol allows simple implementation and goodperformance in small amount of tags, we consider how to takethe advantages of each algorithm. This paper presents query treebased reservation for efficient RFID tag anti-collision. Accordingto the simulation results, the proposed protocol requires lesstime consumption for tag identification than the present tag anti-collision protocols implemented in EPC Class 0, Class 1, andClass 1 Gen. 2.

    Index Terms Anti-collision protocols, distributed reservation-based protocols, tag collisions, RFID.

    I. INTRODUCTION

    RFID systems are coming the most promising technolo-gies used for contactless object identification. One ofthe research areas in RFID systems is a tag anti-collision

    protocol; how to reduce identification time with a given

    number of tags in the field of an RFID reader [1]. For

    fast tag identification, anti-collision protocols, which reduce

    the frequency of collisions occurred and identify tags inde-

    pendent of occurring collisions, are required. Moreover, tag

    anti-collision protocols are more important than reader anti-

    collision protocols because of the incompetence of tags in

    low-cost passive RFID systems.

    There are two types of tag anti-collision protocols for RFIDsystems: deterministic methods and probabilistic methods [1].

    The deterministic methods are tree based protocols [1,2,3,6].

    The tree based protocols keep splitting the group of colliding

    tags into two subgroups until all tags are identified. The

    probabilistic methods are based on slotted ALOHA [4]. The

    slotted ALOHA based protocols reduce the probability of

    occurring tag collisions how tags respond at the different time.

    This paper considers how to improve the performance of

    the RFID tag anti-collision protocol by applying the char-

    acteristics of EPC Class 1 Gen. 2 protocol to the query

    tree algorithm. The crux of the proposed query tree based

    reservation is that the RFID systems do not have to use thewhole length of tag IDs for resolving each tag in the field of

    a reader. Thus, the tags generate reservation sequences, which

    are 16-bit random numbers (RN16s), for assigning slots to

    transmit their IDs, and the reader identifies tags with the query

    tree algorithm using the RN16s. During the identification

    process, the reader calls a tag with the received RN16 for

    collecting the tag ID when a RN16 is identified by the reader.

    Manuscript received September 12, 2006. The associate editor coordinatingthe review of this letter and approving it for publication was Dr. FabrizioGranelli. This work is supported by the RFID/USN standardization project ofETRI under the information communication standardization program of MIC

    (Ministry of Information and Communication), Korea.J. H. Choi, D. Lee, and H. Lee are with the School of Engineering,Information and Communications University (ICU), Daejeon, Korea (e-mail:{jihwanchoi, dwook, hjlee}@icu.ac.kr).

    Digital Object Identifier 10.1109/LCOMM.2007.061471.

    Hence, we call the proposed protocol as 16-bit random number

    aided query tree algorithm (RN16QTA). According to the

    simulation results, RN16QTA achieves less time consumption

    for the tag identification than the present tag anti-collision

    protocols implemented in EPC Class 0, Class 1, and Class 1

    Gen. 2.

    I I . QUERYT RE E-BASEDR ESERVATION(RN16QTA)

    Since the slotted ALOHA based algorithms cannot guaran-

    tee the100%read rate, and experience significant performancedegradation in the large amount of tags, this paper considers

    the performance enhancement of the tree based algorithm. Forenhancing the performance of the tree based algorithm, we

    apply the characteristics of EPC Class 1 Gen. 2 protocol to

    the query tree algorithm as follows:

    1) RN16 Generation: All tags in the field of a reader gen-

    erate temporary IDs, 16-bit random numbers (RN16s),

    for giving the uniqueness to themselves [4]. This part is

    performed exactly like in standard EPC Class 1 Gen. 2.

    Probability of a single RN16:

    0.8/216 < P(RN16 = j) < 1.25/216, where jis any possible number generated by the random

    number generator (RNG). Probability of simultaneously identical sequences:

    For a tag population up to 1,000 tags, the probability

    of existing tags having the same RN16 is less than

    0.1%. Probability of predicting a RN16:

    A RN16 shall not be predictable with a probability

    greater than 0.025%.

    2) Applying RN16s to the Query Tree Algorithm (QTA

    [3])

    Request: The reader sends n-length inquiring bits

    (prefix) to tags. Response: Tags send their RN16s from n+ 1th bit

    to the end when the first n bits of the tag IDs are

    the same as the prefix.

    Decision: Depending on whether collisions have

    occurred or not, the reader decides on proceeding

    procedure with the following conditions.

    If collisions occur, the reader saves two new

    prefixes at the last input first output (LIFO).

    Two new prefixes: the connection of the prefix

    and either 0 or 1

    If a collision occurs when the tags respond withonly the last bits in the RN16s to the readers

    inquiring, the reader assumes that there are two

    tags because of the uniqueness of RN16s.

    1089-7798/07$20.00 2007 IEEE

  • 8/12/2019 04114235

    2/3

    86 IEEE COMMUNICATIONS LETTERS, VOL. 11, NO. 1, JANUARY 2007

    Initialization

    A reade rs Req uest

    (n-bit) Prefix=LIFO (end, :)

    LIFO=LIFO(1:end-1, :)

    For all tags

    TmpID(1:n)=Prefix

    Tags Response from all activated tags

    Res=TmpID(n+1: end)

    Is there any Response?

    1LIFO=

    0

    Inquiring bits

    Answer bits

    LIFO

    LIFO= Prefix 1

    Prefix 0

    Length(LIFO) =0

    End

    No

    Y es

    Y es

    No

    No: Inactivate tags

    Yes: Activate tags

    Does a collision occur?

    Detect_ID

    The received IDDetect_ID=

    Y es

    No

    Temporary IDs

    Each tag generates

    its own 16 bit random number (RN16)

    ACK signal

    The reader send the ACK signal to make

    the tag having the received RN16 send its tag ID

    Detect_RN16=[Prefix Res]

    Tags ID Response

    The tag, which responds to the readers

    inquiring a moment ago, sends its tag ID

    :The process in the reader

    :The process in the tag

    [ ]

    Fig. 1. The flow chart of query tree-based reservation.

    If there is no collision, the reader identifies a

    RN16 from the tags.

    Do these steps until a RN16 is identified.

    3) Sending an ACK command when a RN16 is identified

    If the reader identifies a RN16, the reader calls the

    tag having the RN16 with an ACK command. Thereare two types of the ACK command.

    The reader sends a two-bit ACK code only for

    indicating that the signal is an ACK command.

    The reader sends the ACK code + the received

    RN16 for improving the security of the RFID

    systems.

    Only the tag sending the whole RN16 is activated

    by the ACK command of the reader, and responds

    with its tag ID to the ACK command.

    4) Do (2) and (3) until the LIFO is null.

    The performance of the RFID tag anti-collision protocols

    means how fast the identification process in the field of

    a reader is. In the tree based algorithms, the performance

    depends on the tree depth, which is equal to the length of the

    tag IDs. Here, we propose the query tree based reservation

    called RN16QTA for efficient RFID tag anti-collision as

    shown in Fig.1. RN16QTA does the identification process with

    temporary IDs shortening the tree depth. After identifying a

    temporary ID, the reader calls a tag with the temporary ID for

    collecting the tag ID. With RN16QTA, consequently, we can

    reduce the identification time by the shortened tree depth.

    III. SIMULATIONS ANDR ESULTS

    This simulation is intended for the performance compari-

    son between the present tree based RFID tag anti-collision

    protocols and RN16QTA in the algorithm point, and the per-

    formance comparison between RN16QTA and DFS-ALOHA

    in EPC Class 1 Gen. 2 protocol in the implementation point.

    One important point of the observation is whether the

    simulated protocol is implementable to the low-cost passive

    RFID systems or not. QTA is already applied to the RFID

    systems with EPC Class 1 protocol. On the other hand,

    collision tracking tree algorithm (CTTA [3]) is not adopted toany RFID systems yet because tags cannot support CTTA with

    transmitting and receiving simultaneously. Since RN16QTA is

    the protocol applying the characteristics of DFS-ALOHA in

    EPC Class 1 Gen. 2 protocol to the query tree algorithm,

    the algorithm is easily implementable for the low-cost passive

    RFID systems.

    A. Simulation environment

    The simulation condition is as follows. There is only one

    reader. In the field of the reader, the number of tags increases

    from 2 to 512. The length of the tag IDs is 96 bits. Both tag-to-reader data rate and reader-to-tag data rate are chosen as

    80 kbps. The reason is that the middle speed in EPC Class

    1 Gen. 2 proposed by EPCglobal to ISO/IEC 18000-6 C is

    equal to the chosen data rate [4,5]. There is some iteration

    overhead because of propagation delay from the channel and

    latency from the signal processing. For the comparison of the

    algorithm point, the iteration overhead is not considered.

    For the comparison of the implementation point, we set

    a simple EPC Class 1 Gen. 2 architecture without consid-

    ering both the multi-session commands for the multi-reader

    environment and the probability of receiving an invalid tag

    ID. To calculate the average number of identified tags persecond, the RFID systems choose 8 bits for request and 3

    bits for detecting collision, no collision, or no response for

    substituting the iteration overhead. For solving error detecting

    problem, moreover, 16-bit CRC coding is required.

    B. Results

    1) The algorithm point: Above all else, Fig.2 shows the

    average inquiring bits and response bits for one-tag identifi-

    cation. In the RN16QTA, both the average inquiring bits and

    response bits for one-tag identification are between the case

    of QTA and the case of CTTA. The reason is that RN16QTAreduces the overhead from both the average inquiring bits and

    response bits by using the temporary IDs, but generates the

    overhead caused by collecting the tag IDs. The former one

    affects RN16QTA to get better performance than QTA, and the

  • 8/12/2019 04114235

    3/3

    CHOI et al.: QUERY TREE-BASED RESERVATION FOR EFFICIENT RFID TAG ANTI-COLLISION 87

    101

    102

    0

    50

    100

    150

    200

    250

    The number of tags

    Theaveragerequiredb

    itsforone-tagidentification

    Inquiring - QTA

    Inquiring - CTTA

    Inquiring - R N16QTA

    Response

    Response

    Response

    - QTA

    - CTTA

    - RN16QTA

    Fig. 2. The average inquiring and response bits for one-tag identification.

    101

    102

    0

    0.5

    1

    1.5

    2

    2.5

    3

    3.5

    4

    4.5

    5

    The number of tags

    Theaveragerequirediterationsforone-tagiden

    tification

    QT A

    CTTA

    RN16QTA

    Fig. 3. The average required iterations for one-tag identification.

    latter one influences RN16QTA to get worse performance than

    CTTA. Next, Fig.3 indicates the average required iterations for

    one-tag identification. RN16QTA requires one more iteration

    than QTA because of the step collecting the real tag IDs,

    but the overhead from each iteration is 1/6 of the others.

    Finally, Fig.4 shows the average number of identified tags

    per second in each algorithm. In conformity with the simula-

    tions, RN16QTA achieves faster identification than QTA, and

    approaches to the performance of CTTA.

    2) The implementation point: Fig.5 indicates the averagenumber of identified tags per second in each protocol. De-

    pending on the default frame size (16 slots or 64 slots) in EPC

    Class 1 Gen. 2 protocol, the tendency of the performance is

    different in the small amount of tags. However, the perfor-

    mance of DFS-ALOHA in EPC Class 1 Gen. 2 protocol is

    always less than that of the proposed RN16QTA as shown

    in Fig.5. Consequently, the proposed protocol, RN16QTA,

    accomplishes faster tag identification than the implemented

    protocols, QTA in EPC Class 1 protocol and DFS-ALOHA in

    EPC Class 1 Gen. 2 protocol.

    IV. CONCLUSION

    This paper proposes the query tree based reservation, which

    reduces the tree depth on the identification process for decreas-

    ing the identification time. The query tree based reservation

    101

    102

    0

    100

    200

    300

    400

    500

    600

    700

    800

    900

    1000

    The number of tags

    Theaveragenumberofidentifiedtagspersecond

    QT A

    CTTA

    RN16QTA

    Fig. 4. The average number of identified tags per second.

    101

    1020

    50

    100

    150

    200

    250

    300

    350

    400

    450

    500

    The number of tags

    Theaveragenumberofidentifiedtagspersecond

    Gen.2 :16s lots

    Gen.2 :64s lots

    RN16QTA

    Fig. 5. The average number of identified tags per second.

    applies 16-bit random numbers to the query tree algorithm,

    which is for assigning slots to transmit tag IDs, instead of the

    real tag IDs. According to the simulation results, the query tree

    based reservation achieves less time consumption for the tag

    identification than the implemented tag anti-collision protocols

    in EPC Class 0, Class 1, and Class 1 Gen. 2.

    REFERENCES

    [1] J. Myung, W. Lee, and J. Srivastava, Adaptive binary splitting forefficient RFID tag anti-collision, IEEE Commun. Lett., vol. 10, no. 3,pp. 144-146, Mar. 2006.

    [2] T. Wang, Enhanced binary search with cut-through operation for anti-collision in RFID systems, IEEE Commun. Lett., vol. 10, no. 4, pp.236-238, Apr. 2006.

    [3] F. Zhou, D. Jin, C. Huang, and M. Hao, Optimize the power consumptionof passive electronic tags for anti-collision schemes, in Proc. Fifth Inter.Conf. on ASIC 2003, vol. 2, pp. 1213-1217.

    [4] EPCTM radio-frequency identification protocols class-1 generation-2UHF RFID protocol for communications at 860MHz-960MHz Ver. 1.0.9,EPCglobal, Jan. 2005.

    [5] Information technology automatic identification and data capturetechniquesradio frequency identification for item management air

    interfacepart 6: parameters for air interface communications at 860-960MHz, ISO/IEC FDIS 18000-6, Nov. 2003.[6] J. I. Capetanakis, Tree algorithms for packet broadcast channels, IEEE

    Trans. Inf. Theory, vol. 25, pp. 505-515, Sept. 1979.