Beruflich Dokumente
Kultur Dokumente
=
=
n
p p
n
p
succ idle coll
L L
n
L
P P P
(1)
Where n and L
p
are the number of tags and the pilot frame
size. P
coll
, P
idle
and P
succ
are the probability of collision, idle
and success, respectively.
In the beginning, to decrease the number of timeslots
required for tag estimation, the tags to respond to a reader are
grouped into M subgroups by using bit masks. After this step,
the pilot frame (L
p
) that estimates the number of tags is
applied to only the first subgroup. The other subgroups
predict the number of tags without using L
p
, assuming
uniform random distribution of tags. Even if bit masks may
not have completely uniform random distribution, predicting
the number of tags via bit masks can save more timeslots
because of not using timeslots for tag estimation. If, however,
M is too large, tags among subgroups may not almost have
uniform distribution. So, an appropriate value of M is decided
via simulations. A reader assigns additional timeslots at the
end of L
p
to resolve collisions occurred in L
p
. The tags
collided in L
p
come to have new random counter values
depending on remaining timeslots in L
p
and collision number
in L
p
so that they are resolved after the end of L
p
. The FSAPB
algorithm can be classified into low and high collision
probability cases. During L
p
, all the tags in a subgroup
transmit their IDs. After L
p
is over, a reader calculates P
coll
,
which is compared with the threshold P
th
.
A. Low Collision Probabilitv
When P
coll
is lower than P
th
, binary tree algorithm is
applied since the number of tags collided in a timeslot is
relatively small. Thus, to resolve tags collided in L
p
, a reader
use additional timeslots L
add
.
Fig. 4 shows an example of the proposed FSAPB algorithm,
when P
coll
is lower than P
th
. Tags send their IDs to the reader
when the random counter values are 0. The reader can identify
tag 1 and tag 7 in L
p
as they transmit successfully. Tag 2, 3
and 5 cause the first collision in L
p
. So, the reader assigns two
additional timeslots after L
p
and those tags conduct binary tree
algorithm during L
add
. Tag 4 and 6 select random counter
values 0 or 1 by binary tree algorithm and add them by the
two times the number of collisions before tag 4 and 6 to the
current counter values by the second collision in L
p
. Tag 2 and
tag 5 collide again in the first timeslot of L
add
, and binary tree
algorithm is conducted. So, the other tags in L
add
increase the
counter values by 1.
B. High Collision Probabilitv
If there are many tags, most of timeslots in L
p
may
experience collisions. Owing to these collisions, P
coll
in L
p
2007 International Symposium on Communications and Information Technologies (ISCIT 2007) 1029
Figure 5. An example of FSAPB algorithm, when Pcoll is larger than Pth. Figure 6. When n=100, the number of required timeslots vs varying
frame size.
may be greater than P
th
. In this case, the number of tags
collided in a timeslot becomes large. Thus, L
add
is not used,
but a new frame L
1
is used for tag identification. The size of
L
1
can be calculated by letting L
1
n from Eq. (1). It is not
necessary to estimate the number of tags (n) accurately, since
many timeslots for accurate tag estimation may be wasted.
Considering the succeeded tags in L
p
and the optimal frame
size, the size of L
1
is set by the minimum value of n that meets
Eq. (1). The optimal frame size is referred to in the next
section.
Fig. 5 shows an example of the proposed FSAPB algorithm,
when P
coll
is greater than P
th
. After L
p
, another frame L
1
is
created. All the tags collided in L
p
retransmit their IDs in L
1
.
During the progress in L
1
, the third timeslot has collision by
tag 4 and tag 6. So, the other tags not succeeded in L
1
increase
their random counter values by 1. This step is repeated for
other timeslots.
A reader can know the number of tags of the other
subgroups after it identifies the tags in the first subgroup,
because bit masks separate tags uniformly. The suitable frame
size of the current subgroup is determined by multiplying a
certain constant (y) by the number of tags identified in the
previous subgroup. In the next section, we show why a certain
constant has to be multiplied. After the suitable frame size is
set, binary tree protocol is conducted.
Figure 7. 1/Lopt vs the varying number of tags.
C. The Optimal Frame Si:e
In DFSA, when the number of tags equals to the frame size,
it is optimal. However, Fig. 6 shows FSAPB is not optimal
when the number of tags equals to the frame size. Thus,
FSAPB requires a new optimal frame size (L
opt
). Let T, n and
k be the total number of timeslots used for tag identification,
the total number of tags and the number of tags transmitted in
a timeslot, respectively.
(2)
Where o
k
is the average number of timeslots used by binary
tree when collisions occur [6]. We can find the optimal frame
size by differentiating Eq. (2). Although it is very difficult to
differentiate Eq. (2) for the various n, we can obtain that the
L
opt
is about 0.88n from Fig. 7. It shows that the graph is not
affected by k when k is large. Hence, we can differentiate Eq.
(2) in the special range of k. That is the reason why a certain
constant (y) in the previous section is multiplied.
IV. SIMULATION RESULT
We evaluate the performance of FSAPB compared to that
of DFSA, binary tree and EB-FSA algorithms under varying
number of tags. Our simulation condition is as follows. Only
one reader attempts to read all the tags. The number of tags n
increases from 100 to 1000. In this simulation, FSAPB is set
to L
p
=36, P
th
=0.6 and M = 4, and the simulation of EB-FSA
is referred to [7]. In DFSA, we assume the initial number of
tags is known and tag estimation is very accurate.
. ) 1 (
1
1
1
0
=
n
k
k opt
k n
opt
k
opt
k n
L
L L
C T
1030 2007 International Symposium on Communications and Information Technologies (ISCIT 2007)
Figure 8. The number of slots wasted for tag estimation with the varying
number of tags.
Fig. 8 shows the number of timeslots wasted for tag
estimation by varying the number of tags, and FSAPB
consumes less timeslots than EBFSA for tags estimation.
When the number of tags n is less than 200, P
coll
in L
p
is lower
than threshold. So, a reader uses L
add
, which has good
performance by binary tree protocol due to the small number
of tags in L
p
. When, however, it is from n=300 to n=1000, L
1
is used for tag identification instead of L
add
owing to a large
collision in L
p
. That is the reason why the FSAPB graph in
Fig. 8 demonstrates some discontinuity. And, FSAPB does
not consume more timeslots than L
p
. EBFSA, however,
requires many timeslots as the number of tags is large. When
n=1000, EBFSA requires five and a half times as many
timeslots as our FSAPB.
We compare FSAPB with ALOHA based protocol and
binary tree protocol. Fig. 9 depicts the number of timeslots
used to identify tags with varying number of tags. We can
observe that most of the algorithms have nearly the same
performance when the number of tags is small. As the number
of tags becomes greater, FSAPB is more efficient than the
other algorithms. When the number of tags is 1000, EBFSA,
DFSA and binary tree consume more timeslots about 7.3%,
14.7% and 21.9% than FSAPB.
Figure 9. The number of slots used to identify tags with the varying
number of tags.
V. CONCLUSION
We have proposed tag estimation by the pilot frame and tag
identification by both DFSA and binary splitting. Proposed
tag estimation method using the pilot frame saves the number
of timeslots required for tag estimation. Moreover, in tag
identification, performance of FSAPB is superior over that of
EBFSA because FSAPB has the near optimal frame size. And
simulation results show that FSAPB consumes less timeslots
than EBFSA, DFSA and binary tree protocol.
REFERENCES
[1] R. Want, An Introduction to RFID Technology, IEEE
Pervasive Computing, vol. 5, no. 1, pp. 25-33, Jan. 2006.
[2] EPCTM Radio-frequency Identification Protocols Class-1
Generation-2 UHF RFID Protocol For Communications at
860MHz-960MHz Version 1.0.9, EPCglobal, Jan. 2005.
[3] J. Myung, W. Lee and J. Srivastava, Adaptive Binary Splitting
for Efficient RFID Tag Anti-Collision, IEEE Comm. Letters,
vol. 10, no.3, pp. 144-146, Mar. 2006.
[4] S. Lee, S. Joo and C. Lee, An Enhanced Dynamic Framed
Slotted ALOHA Algorithm for RFID Tag Identification, in
Proc. of MobiQuitous, pp. 166-172, Jul. 2005.
[5] J. Cha and J. Kim, Novel Anti-collision Algorithms for Fast
Object Identification in RFID System, in Proc. of Parallel and
Distributed System, vol. 2, pp. 63-67, Jul. 2005.
[6] A. J. E. M. Janssen and M. J. M. de Jong, Analysis of
Contention Tree Algorithms, IEEE Trans. on Information
Theory, vol. 46, no. 6 pp. 2163-2172, Sep. 2000.
[7] J. Park, M. Chung, and T.-J. Lee, Identification of RFID Tags
in Framed-Slotted ALOHA with Robust Estimation and Binary
Selection, IEEE Communications Letters, Vol. 11, no.5, pp.
452-454, 2007.
2007 International Symposium on Communications and Information Technologies (ISCIT 2007) 1031