Vous êtes sur la page 1sur 6

A Novel Time- and Energy-Efficient Tag

Identification Protocol in Dense RFID Systems


Linh T. Hoang∗ , Chuyen T. Nguyen∗ , and Anh T. Pham†
∗ School of Electronics and Telecommunications, Hanoi University of Science and Technology, Vietnam
E-mail: linhhoangtuan95@gmail.com, chuyen.nguyenthanh@hust.edu.vn
† Computer Communications Lab., The University of Aizu, Japan
E-mail: pham@u-aizu.ac.jp

Abstract—This paper proposes a novel time- and energy- protocols continuously split tags into multiple groups until
efficient identification protocol in dense radio frequency identifi- there is at most one tag in each group. Therefore, all tags
cation (RFID) systems. The protocol is based on the conventional can be recognized within a certain time. Tree-based protocols
M-ary collision tree (MCT) where tags involving a collision are
classified into other M sub-trees. Nevertheless, different from are expanded into two classes i.e., binary tree (BT) and query
the MCT, we design a transmission mechanism by which each tree (QT) depending on the splitting mechanism. While QT
tag only responds to the reader by a small number of bits for a uses tags’ IDs, BT uses random numbers for the splitting
collision detection. The mechanism is relied on a collision window mechanism, as a result, memory is needed in tags. The QT
and Manchester encoding that are widely used for RFID systems. protocols are therefore more practical and are in the focus of
Thanks to the mechanism, the number of bits transmitted by
tags is significantly reduced, which improves the overall system this paper.
performance in terms of both time and energy consumption. In the evaluation of RFID protocols, both identification
Performance analysis is investigated to validate the correctness of time and energy consumption are two important performance
the mechanism. Computer simulations are also performed using metrics. Most of early works, however, focus on optimizing the
Monte-Carlo method. The obtained results are then compared identification time, i.e. “how to rapidly identify all the tags”.
with those of the MCT to confirm the effectiveness of the proposed
protocol. The most time-efficient protocols are the bit-tracking QT-
Index Terms—Dense RFID, identification, time-efficient, based ones, including optimal query tracking tree (OQTT) [7],
energy-efficient, Manchester encoding collision tree (CT) [8], dual prefix probe (DPPS) [9], and the
improved assigned tree slotted aloha (ImATSA) [10]. In these
I. Introduction protocols, Manchester coding is used to encode each tag’s
Radio frequency identification (RFID) has become one of response by which the position of colliding bits (if any) can be
the best known technologies in identifying objects automat- detected. Thanks to this capability, the reader can utilize the
ically through radio frequency (RF) channel for years. The first position to generate a “common prefix” of corresponding
technology also holds a key role in the picture of the Internet of colliding tags. Those tags are then split into smaller groups
Things (IoT) where millions of objects are employed and mon- with new prefixes by appending one bit (OPTT, CT) or more
itored [1], [2]. A typical RFID-based system includes a reader bits (DPPS, ImATSA) to the common one. Nevertheless, since
and a number of tags, where each tag has a unique identity (ID) only the first colliding bit is used, many collisions are still
[3]. The reader tries to identify all the IDs efficiently in terms generated.
of time and/or energy consumption. Nevertheless, one of the Regarding the energy consumption in passive RFID sys-
main challenges that severely affects to the ID identification tems, it consists of two portions: one for powering tags, and
is the tag collision [1]. It happens when more than one tags the other is consumed by the reader to send/receive messages
transmit their signal to the reader simultaneously. Due to the from tags. While the first portion is proportional to the
shared RF channel, the reader can not decode the transmitted identification time, the second one depends on the number of
signal. Therefore, other retransmissions are required, which bits that the reader transmits and receives. Therefore, in order
results in the inefficiency of system performance, especially to optimize the energy consumption, both the identification
when the number of tags is large. time and the number of transmitted bits should be as low as
To cope with the tag collision problem, a number of possible [11]. Recently, the M-ary collision tree (MCT) has
identification protocols have been proposed. They are mainly been proposed taking into account both identification time
based on two different approaches i.e., tree-based [4] and and energy consumption performance metrics [12]. In this
aloha-based [5]. In aloha-based protocols, the identification protocol, the first log2 M colliding bits are utilized to split
process is separated into multiple frames of time slots, and involving tags into M smaller groups. With the information
each tag randomly responds in one of the slots. Although of more colliding bits, the overall identification time and the
the protocols are simple, there is no guarantee on the time number of transmitted bits from tags are saved; and it is seen
required to read all tags [6]. On the other hand, tree-based that more than 15% of both time and energy consumption
can be improved in comparison with conventional protocols. TABLE I
Nevertheless, the number of bits that the reader receives for Symbol Definitions
collision detection is still large, especially in dense systems. Symbol Definition
Therefore, we believe that there is room for MCT performance Prefix (of l bits length) transmitted by the reader
pre
improvement. at the beginning of each frame.
In this paper, a novel time- and energy-efficient identifi- Bit position of the i-th colliding bit in pre, where
bpi
i ∈ [1, log2 M − 1].
cation protocol is proposed adopting key features of MCT. The window bit in each query. The window is
wd
Furthermore, a transmission mechanism is designed in the active if wd = 1 and deactivated if wd = 0.
protocol in which each tag only responds to the reader by Matching prefix used to control tag replies. It is
mPre obtained by deleting all the colliding bits (e.g., the
a very small number of bits for the detection of colliding bits. bpi -th bit) in pre.
The mechanism is relied on a collision window proposed in Tag’s matching ID, which is obtained by deleting
tID
[13] where the number of transmitted bits is required within all the bpi -th bit in ID(1 : l).
Query queue maintained by reader to record the
a predefined window. Thanks to the mechanism, the total Q
frame parameters.
number of bits the reader receives is significantly reduced, Frame index i.e., the i-th frame in the recognition
Fi
which improves MCT performance in terms of both time and process.
energy consumption. Performance analysis is investigated to Slot index i.e., the (x + 1)-th slot in the current
Sx
frame, where x ∈ [0, M − 1].
validate the correctness of the mechanism. Computer simu- The binary form of each slot index, S b =
Sb
lations are also performed using Monte-Carlo method. The de2bi(x, log2 M), where x ∈ [0, M − 1].
obtained results are then compared with those of the MCT to comm The common prefix of each slot.
DM Decoded received message at the reader.
confirm the effectiveness of the proposed protocol. The operation removing the first element from Q
QueueOut(Q)
after it is broadcasted in MCT protocol.
II. System Model The operation to insert the new element denoted
QueueIn(Q, ele)
A. System Description by ele into Q.

The considered dense passive RFID system consists of a


reader and a large number of passive tags, denoted by n.
B. Time and Energy Models
Each tag is represented by a unique 128-bit identity (ID). The
communication between the reader and tags is in half-duplex According to the link timing in Fig. 1, time and energy
mode, while transmission channels between them are assumed models are described as follows:
error-free for the sake of simplicity. Time Model: Let’s first define T (n) as the execution time
The reader aims to efficiently collect all the IDs in terms to identify the n tags. It consists of time for reader’s request
of time and energy consumption using query tree-based pro- commands, tags’ responses, and waiting time denoted by T req ,
tocols. In the tree-based protocols, a query is broadcasted T res , and T wait , respectively. T req includes time for transmitting
by the reader to ask for tags’ reply. If a tag’s ID matches Query at the i-th frame and Qrep at the beginning of each time
with the so-called prefix in the query message, it responds slot of the frame, which are denoted by tQi and tR , respectively.
to the reader. In other cases, it just keeps silent. During the For tags’ responses, we define tT j as the time for transmitting
transmission, if the query results in a collision i.e., several tags’ replies in the j-th non-empty (collision or success) slot.
tags simultaneously reply, the first log2 M colliding bits (the Then, T (n) can be written by
mechanism for colliding bit detection is further explained) are T (n) =T req + T res + T wait . (1)
used to split involving tags into M smaller groups. Moreover,
depending on the number of simultaneous responding tags, Energy Model: During the identification process, the reader
there are three types of time slot namely, collision, empty, needs to broadcast the continuous waves (CWs) with a power
and success. In particular, there is no tag in each empty slot, of Ptx to provide energy for passive tags. On the other hand,
while in success slot, there is only one and the reader can the reader needs the extra power of Prx during the tags’
successfully identifies the tag’s ID. In the other cases, the slot transmitting period. Ptx and Prx are determined in Joule per
is collision. second (J/s) obtained from [6]. Then, to collect n tags, the
To better describe the protocol, we illustrate in Fig. 1 the total energy consumption denoted by E(n) is given by
link timing between the reader and tags. The identification S (n)−C
Xe (n)
process includes multiple (M-slot) frames. Each frame starting E(n) = Ptx T (n) + Prx tT j , (2)
with a Query command broadcasted by the reader in time of j=1
tQ is followed by M continuous slots. Each slot starts with the where S (n) and Ce (n) are denoted as the total number of slots
reader’s Qrep command in time of tR except the first one. tT and the number of empty slots, respectively.
is the time needed for a tag’s response. t1 is the time taken
for signal transmission from the reader to tags, while t2 is III. Proposed Protocol
the time for backscattering the signal to the reader. Finally, t3 In this section, we first present a conventional M-ary
is the waiting time of the reader after t1 if there is no tag’s collision tree (MCT) protocol [12] that shares the same as-
response. sumptions of the transmission models and systems as ours.
Success Slot Collision Slot Empty Slot

Reader Query CW Qrep CW Qrep CW Qrep CW Query

No At least one No
colliding bit colliding bit response
Tag Response Response
tQ t1 t2 tR t1 t2 tR t1 t3
tT tT
Frame (M slots)

Fig. 1. Link timing between the reader and tags with collision/empty/success slots

Based on key features of MCT, we then propose another M- ID(5,6). In this case, tag C transmits “01” at slot S 4 of this
ary collision window tree protocol (MCwT), which is expected frame (since ID(bp1 , l + 1)=“11”) where it is successfully
to improve the MCT performance in terms of both time and decoded. Tags A and B with transmitted bits of “11” and
energy consumptions. For convenience, we show in Table I “00”, respectively, result in a collision at slot S 1 of the frame
several symbols and procedures used for both the protocols. due to ID(bp1 , l + 1)=“00”. Besides, it is noted that for the
Note that bp indicates the bp-th bit in each prefix pre that first query all the log2 (M) − 1 bits in the pre are marked as
is colliding, and is not used for prefix matching. de2bi(x, k) colliding, and thus mPre = tID = ⊘ (empty string). In this
converts a decimal number x into k-bit binary string. mPre case, all the tags respond with the first log2 (M) ID bits. Note
and tID are parameters used to control tags’ responses. If a that, str(i) refers to the i-th bit of the string str and str(i : j)
tag has mPre = tID, it will reply to the reader for the current represents for the bit string from str(i) to str( j). Note that
query. Moreover, the reader adopts a first-in-first-out query j =“end” indicates the last bit of the string str, while str(i : j)
queue denoted by Q. Each element of the queue is used for refers to an empty string if i > j. str1 ||str2 concatenates strings
a query and then is removed through QueueOut(Q), while str1 and str2 .
another element is inserted into the queue if a collision occurs
New prefix composing at the reader side: To better explain
via QueueIn(Q, ele). The process continues until Q becomes
the prefix composing process, we denote by comm the com-
empty.
mon prefix at each slot where it is set as comm=“pre(1 : bp1 −
1)||S b (1)||pre(bp1 + 1 : bp2 − 1)||S b (2)||...||pre(bplog2 (M)−1 − 1 :
A. The conventional MCT protocol
l)||S b (log2 M)”0 . For example, the prefixes of slots S 1 and S 4
In MCT, each element in Q consists of a bit string of pre of the second frame are, respectively, “0000” and “0011”. If
and values of bpi for i = 1, ..., log2 (M) − 1. The initial queue no colliding bits are detected at a received message DM, the
is Q = {“11
| {z. . . 1”
}; 1, 2, . . . , log2 (M) − 1}. When a collision slot reader can obtain the involving tag’s ID as comm||DM. For
log2 (M)−1 example, in the slot S 4 the decoded DM =“01” and thus,
occurs, the reader records positions of those first log2 (M) − 1 the ID of tag C is successfully identified as “001101”. On
colliding bits to split the colliding tags into M subgroups. In the other hand, if a collision happens the reader can detect
other words, the identification process is divided into multiple the position of the i-th colliding bit (i ∈ [1, log2 M]) in the
M-slot frames and each group corresponds to a slot. The per- corresponding DM denoted as Ci thanks to Manchester coding.
formance of MCT is explained in more details at tag and reader Then, the colliding bit is replaced by bit “1” and a new prefix
sides via a simple example with 6 tags A, B, C, D, E, and is generate by pre =“comm||DM(1 : C1 − 1)||1||DM(C1 + 1 :
F as in Fig. 2 and Table II as follows. Note that the state of C2 − 1)||1||...||DM(Clog2 (M)−1 + 1 : Clog2 M − 1)”. The prefix is
each slot is denoted by one of the symbols C, E, S, G, which inserted into Q for the query process. In slot S 1 of F2 , the
refer to a collision, empty, success, ongoing slot, respectively. received DM is “??” and the new pre in F3 is, therefore,
Prefix matching at the tag side: After receiving “00001”. The reader repeats frames until Q becomes empty.
Query(pre, bp1 , . . . , bplog2 (M)−1 ), each tag first calculates Based on the above performance of the MCT, the identification
mPre and tID by deleting all the bpi -th bits in pre and time T (n) is also found as
ID(1 : l), respectively, where l is the length of the pre. Taking [Cc (n)+CeX
(n)+C s (n)]/M Cc (n)+C
X s (n)
an example for the second query where (pre, bp)=(“001”, 3) T (n) = [tQi + (M − 1)tR ] + tT j
(M = 4 and l = 3 in this case) to handle the collision caused i=1 j=1
by tags A, B, and C. Here, mPre =“00” and tID = ID(1, 2). + [(t1 + t2 ) (Cc (n) + C s (n)) + (t1 + t3 )Ce (n)] , (3)
Then, if tID does not match with mPre, the tag keeps silent.
Otherwise, the tag converts its ID(bp1 , . . . , bplog2 (M)−1 , l + 1) where the first, second, and third terms of (3), respectively,
to a slot index S x and it transmits the rest of its ID, i.e. refer to T req , T res , and T wait in (1). Cc (n) and C s (n) are
ID(l + 2 : end), in the (x + 1)-th slot. In our example, tags denoted by the total numbers of collision and success slots,
A, B, and C have mPre = tID =“00” so that they transmits respectively.
C

F1: (pre, bp) = (“1”, 1)


S1 S2 S3 S4
(A, B, C) (D, E) (F)
C E C S
Tag ID
F2: (pre, bp) = (“001”, 3)
A 0 00 01 1 S1 S2 S3 S4
B 0 00 00 0 (A, B) (C) C Collision slot
C 0 01 10 1 C E E S
D 1 00 00 1 S Success slot
E 1 00 01 1 F3: (pre, bp) = (“00001”, 5)
S1 S2 S3 S4
F 1 11 00 1 E Empty slot
(B) (A)
S E E S

Fig. 2. An example of MCT protocol with 6 tags A, B, C, D, E, and F where M = 4.

TABLE II
The identification process of MCT protocol used in Fig. 2

Slot 1 Slot 2 Slot 3 Slot 4


Frame Query Query Queue
State Message State Message State Message State Message
1 (“1”, 1) C “????” E – C “00?1” S “1001” Q = {(“001”, 3)}
2 (“001”, 3) C “??” E – E – S “01” Q = {(“00001”, 5)}
3 (“00001”, 5) S “1” E – E – S “1” Q=∅

B. Proposed MCwT protocol words, Lcomm + W < K where Lcomm and K is the lengths of
the common prefix comm and tag’s ID, respectively. After this
1) Protocol Description: The MCwT protocol adopts all
type of slot, the reader sets wd as “0” for the next query. In
the key features of the conventional MCT such as Manchester
this case, the frame in the new query has only one slot, which
encoding, M-ary collision tree structure. Nevertheless, another
can be seen in frames F3 and F4 in Fig. 3, while involving
key point to improve the MCT performance in terms of
tags transmit the rest of its ID.
time and energy consumption is proposed in MCwT. In the
The corresponding execution time of MCwT can be written
following, we describe the MCwT in details via the same
as
above example. The corresponding collision tree as well as the
identification process are also illustrated in Fig. 3 and Table III. T (n) = T req + T res + T wait
The main difference between MCT and MCwT is that a  
Q(n)
X Xg (n)
Q(n)−C  S (n)−C
Xe (n)
collision window used to manage the length of bit strings =   t Qi + (M − 1)tR  + tT j
transmitted by tags. In particular, if tID = mPre, a tag only i=1 i=1 j=1
transmits a few bits within a predefined window size denoted  
+ (t1 + t2 ) Cc (n) + C s (n) + Cg (n) + (t1 + t3 )Ce (n), (4)
by W instead of the rest of its ID. In order to employ the
window, each query from the reader consists of three following where Cg (n) is the total number of ongoing slots, while Q(n)
information: the prefix pre, the position of colliding bits bp, is the total number of queries, i.e., Q(n) = [Cc (n) + Ce (n) +
and a window bit denoted by wd. When wd = 1, the window is C s (n)]/M + Cg (n).
active and tags involving in a collision slot only transmit back On the other hand, the tag’s operations in our protocol as
W bits of its ID. For example, colliding tags in Fig. 3 respond well as MCT (such as bit checking, string composing, and
by only two bits to the reader. On the other hand, if the window number-to-string conversion) are quite simple. Therefore, it is
is deactivated by the reader, i.e. wd = 0, then the tags, as same possible to implement the protocol not only in active RFID
as MCT, transmit the rest of its ID. There are three situations systems but also existing passive ones. Moreover, commercial
that the reader deactivates the window. First, a log2 M colliding passive RFID tags such as EM4305, TRF7960, TRF7964 also
bits are not fully detected. Second, the estimated number of support Manchester coding to detect the position of colliding
colliding tags is less than a threshold, which will be further bits [12], which can also validate the possibility of our protocol
analyzed. Finally, the window size is longer than the length of implementation.
remaining bits in tag’s response. Note that the reader knows the 2) The window size and collision detection probability: The
lengths of each ID and the prefix, so it can calculate the length window size is selected as a minimum value so that the reader
of remaining bits. In our example, the window is deactivated can detect log2 M colliding bits from n tags’ responses for the
in F3 , F4 , and F5 in Fig. 3. splitting process. Then, assuming the uniform distribution of
As a result of window mechanism, a new type of slot, all IDs, the probability a transmitted bit is detected colliding
namely “ongoing slot”, is considered at the reader side. It oc- when n tags respond to the reader, which is denoted by q(n),
curs when the reader successfully decodes the tags’ responses can be written as
2n − 2
but the entire corresponding IDs are still not identified. In other q(n) = . (5)
2n
C

F1: (pre, bp, wd) = (“1”, 1, 1)


S1 S2 S3 S4
(A, B, C) (D, E) (F)
C E G G
Tag ID
F2: (pre, bp, wd) = (“001”, 3, 1) F3: (pre, bp, wd) F4: (pre, bp, wd) C Collision slot
A 0 00 01 1 S1 S2 S3 S4 = (“1000”, 0, 0) = (“1110”, 0, 0)
B 0 00 00 0 (A, B) (C) (D, E) (F) S Success slot
C 0 01 10 1 C E E S C S
D 1 00 00 1
F5: (pre, bp, wd) = (“00001”, 5, 0)
E Empty slot
E 1 00 01 1
F 1 11 00 1 S1 S2 S3 S4
(B) G Ongoing slot
(A)
S E E S

Fig. 3. M-ary collision tree in MCwT protocol with the window size (W) of 2 bits

TABLE III
The identification process of MCwT protocol used in Fig. 3

Slot 1 Slot 2 Slot 3 Slot 4


Frame Query Query Queue
State Msg. State Msg. State Msg. State Msg.
1 (“1”, 1, 1) C “??” E – G “00” G “10” Q = {(“001”, 3, 1), (“1000”, 0, 0), (“1110”, 0, 0)}
2 (“001”, 3, 1) C “??” E – E – S “01” Q = {(“1000”, 0, 0), (“1110”, 0, 0), (“00001”, 5, 0)}
3 (“1000”, 0, 0) C “?1” Q = {(“1110”, 0, 0), (“00001”, 5, 0)}
4 (“1110”, 0, 0) S “01” Q = {(“00001”, 5, 0)}
5 (“00001”, 5, 0) S “1” E – E – S “1” Q=∅

TABLE IV
100
Parameters in the simulation

Parameter Value Parameter Value 80


Qrep 4 bits Dr 160 kbps
Probability (%)

t1 , t2 25 us t3 12.5 us 
 W = 2 (theory)
tQ (Lcmd + Lpre )/Dr tT Lpream + Lres /Dr 60 W = 2 (simulation)
Ptx 825 mW Prx 125 mW W = 4 (theory)
W = 4 (simulation)
40

As a result, the probability that exact log2 M colliding bits are


detected in the received message DM when each of n tags 20
0 5 10 15 20
transmits W bits can be given as Number of tags (n)
log2 M
p(n, W, log2 M) = CW q(n)log2 M (1 − q(n))W−log2 M , (6) Fig. 4. The probability that 2 (M = 4) colliding bits are detected within a
window W for a given number of tags
where p(n, W, log2 M) is the probability. Therefore, the prob-
ability that at least log2 M colliding bits are detected in the
received message DM denoted by P(n, W, M) is given by different system parameters. The number of tags n is con-
sidered from 1000 to 5000 and all their IDs are assumed
logX
2 M−1
uniformly distributed. Other parameters, similar to [12], are
P(n, W, M) = 1 − p(n, W, k) set in Table IV. Dr is the data rate and Lcmd is the overhead
k=0
length of the Query(). Note that in MCwT, the overhead length
logX
2 M−1
k is increased by one bit since an extra bit is used to save the
= 1− CW q(n)k (1 − q(n))W−k . (7)
state of the window, i.e., Lcmd = 62. Lpre and Lres are the length
k=0
in bits of the reader’s prefix and a tag’s response, respectively.
In summary, based on (7) the reader can always choose Lpream is the 9-bit preamble in each tag’s response [12]. The
minimum value of the window size so that log2 M colliding simulation results are obtained by Monte Carlo method with
bits are detected for a given threshold of the probability the number of simulation runs of 1000, and are also compared
P(n, W, M) and n. For example, if M = 4 and n = 8 then with those of the conventional MCT.
99,9998% the reader can detect 2 colliding bits with W = 4. We first validates our analysis of the efficiency of the
collision window in detecting a collision slot via Fig. 4. This
IV. Numerical Results and Discussion figure presents the probability of detecting a collision with
In this section, we evaluate the performance of the proposed respect to a given number of tags, for a given window size. It is
MCwT and conventional MCT via computer simulations with seen that the theoretical result matches with the simulation one
tags in each query is estimated accurately, the number bits
8
MCT transmitted back to the reader for the collision detection can
MCwT-Estimate
Execution time T(n) (s) 7
MCwT-Ideal be controlled more efficiently. The performance of MCwT, in
6 this case, is much more improved.
5
V. Conclusion
4
The issue of tag identification in passive RFID systems has
3
been studied in this work in terms of both time and energy
2 consumption. By taking key features of the conventional MCT
1 and CwT protocols, we proposed a novel time- and energy-
1000 2000 3000 4000 5000
Number of tags (n)
efficient MCwT. A transmission mechanism was designed
using a collision window by which lots of transmitted bits were
Fig. 5. Execution time for all tags identification with 128-bit ID saved. The analysis was investigated to prove the correctness
of the mechanism. Computer simulations were also performed.
The obtained results showed that the proposed MCwT outper-
8 forms the conventional MCT both in terms of identification
Energy consumption E(n) (J)

MCT
7 MCwT-Estimate time and energy cost, especially for dense RFID systems and
MCwT-Ideal accurate tag cardinality estimation schemes.
6

5 Acknowledgment
4 This work is supported by Vietnam Ministry of Science and
3 Technology under the project DTDLCN.36/17.
2
References
1
1000 2000 3000 4000 5000 [1] I. S. Miodrag Bolic, David Simplot-Ryl, RFID Systems: Research Trends
Number of tags (n) and Challenges. John Wiley & Sons, Ltd, 2010.
[2] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and
Fig. 6. Energy consumption for all tags identification with 128-bit ID M. Ayyash, “Internet of things: A survey on enabling technologies,
protocols, and applications,” IEEE Communications Surveys Tutorials,
vol. 17, no. 4, pp. 2347–2376, Fourthquarter 2015.
[3] K. Finkenzeller, RFID handbook: Fundamentals and Applications in
(averaged in 50.000 samples), which proves the correctness of Contactless, Smart Cards, Radio Frequency Identification and Near-
our analysis. Moreover, even for a small number of contention Field Communication. John Wiley & Sons, Ltd, 2010.
[4] C. T. Nguyen, A.-T. H. Bui, V.-D. Nguyen, and A. T. Pham, “Modified
tags, the reader can easily detect a collision in most cases only tree-based identication protocols for solving hidden-tag problem in rfid
with a small window size. For example, for wd = 2, 10 tags systems over fading channels,” IET Communications, vol. 11, no. 7, pp.
mostly result in a collision. 1132–1142, May 2017.
[5] K.-W. Chin and D. Klair, Aloha-Based Protocols. John Wiley & Sons,
We now show in Figs. 5 and 6 the total time and energy Ltd, 2010, pp. 179–202.
consumed in both MCT and MCwT protocols to identify n [6] V. Namboodiri and L. Gao, “Energy-aware tag anticollision protocols for
tags. The window size is set by 4, while the number of rfid systems,” IEEE Transactions on Mobile Computing, vol. 9, no. 1,
pp. 44–59, January 2010.
colliding tags to deactivate the window is set by 8. In MCwT, [7] J. Lin, C. Lai, Y. Lai, L. Hsiao, and H. Chen, “A novel query tree
the total number of tags is estimated during the identification protocol with bit tracking in rfid tag identification,” IEEE Transactions
process, which is, however, not presented here due to space on Mobile Computing, vol. 12, pp. 2063–2075, October 2013.
[8] X. Jia, Q. Feng, and C. Ma, “An efficient anti-collision protocol for rfid
limitation. In the perfect case, the number of tags in each tag identification,” IEEE Communications Letters, vol. 14, no. 11, pp.
time slot is assumed to be known to the reader. We can 1014–1016, November 2010.
see that the performance of the proposed protocol is better [9] J. Su, Z. Sheng, G. Wen, and V. C. M. Leung, “A time efficient tag
identification algorithm using dual prefix probe scheme (dpps),” IEEE
than that of the conventional one. Also, this gap increases Signal Processing Letters, vol. 23, no. 3, pp. 386–389, March 2016.
when the total number of tags increases. The reason is that [10] L. Zhang, J. Zhang, and X. Tang, “Assigned tree slotted aloha rfid tag
for a collision detection, MCwT uses a few bits within a anti-collision protocols,” IEEE Transactions on Wireless Communica-
tions, vol. 12, no. 11, pp. 5493–5505, November 2013.
predefined small window size, while each tag in MCT is [11] S. Zhang, X. Liu, J. Wang, J. Cao, and G. Min, “Energy-efficient active
required to transmit the remaining bits of its ID. As a result, tag searching in large scale rfid systems,” Information Sciences, vol.
although the mechanism might slightly increase the number 317, pp. 143 – 156, 2015.
[12] L. Zhang, W. Xiang, X. Tang, Q. Li, and Q. Yan, “A time- and
of queries because of ongoing slots, the total time and energy energy-aware collision tree protocol for efficient large-scale rfid tag
consumption significantly decrease. Moreover, in dense RFID identification,” IEEE Transactions on Industrial Informatics, 2017.
systems where the number of tags is extremely large, the [Online]. Available: http://dx.doi.org/10.1109/TII.2017.2771772
[13] H. Landaluce, A. Perallos, E. Onieva, L. Arjona, and L. Bengtsson, “An
number of tags involving in each query is also very large. As energy and identification time decreasing procedure for memoryless rfid
a result, MCwT saves much more time and energy to detect tag anticollision protocols,” IEEE Transactions on Wireless Communi-
tags in comparison with MCT. When the number of contention cations, vol. 15, no. 6, pp. 4234–4247, June 2016.

Vous aimerez peut-être aussi