Grouping based radio frequency identification anti-collision protocols for dense internet of things application

International Journal of Electrical and Computer Engineering (IJECE)
Vol. 12, No. 6, December 2022, pp. 5848~5860
ISSN: 2088-8708, DOI: 10.11591/ijece.v12i6.pp5848-5860  5848
Journal homepage: http://ijece.iaescore.com
Grouping based radio frequency identification anti-collision
protocols for dense internet of things application
Nnamdi H. Umelo1
, Nor K. Noordin1
, Mohd Fadlee A. Rasid1
, Tan K. Geok2
, Fazirulhisyam Hashim1
1
Department of Computer and Communication Systems Engineering, Faculty of Engineering, University Putra Malaysia,
Selangor, Malaysia
2
Faculty of Engineering, Multimedia University, Melaka, Malaysia
Article Info ABSTRACT
Article history:
Received Jun 5, 2021
Revised Jun 5, 2022
Accepted Jul 1, 2022
Radio frequency identification (RFID) is an important internet of things
(IoT) enabling technology. In RFIDs collision occur among tags because
tags share communication channel. This is called tag collision problem. The
problem becomes catastrophic when dense population of tags are deployed
like in IoT. Hence, the need to enhance existing dynamic frame slotted
ALOHA (DFSA) based electronic product code (EPC) C1G2 media access
control (MAC) protocol. Firstly, this paper validates through simulation the
DFSA theory that efficiency of the RFID system is maximum when the
number tags is approximately equal to the frame size. Furthermore, literature
review shows tag grouping is becoming popular to improving the efficiency
of the RFID system. This paper analyzes selected grouping-based
algorithms. Their underlining principles are discussed including their tag
estimation methods. The algorithms were implemented in MATLAB while
extensive Monte Carlo simulation was performed to evaluate their strengths
and weaknesses. Results show that with higher tag density, fuzzy C-means
based algorithm (FCMBG) outperformed traditional DFSA by over 40% in
terms of throughput rate. The results also demonstrate FCMBG bettered
other grouping-based algorithms (GB-DFSA and GBSA) whose tag
estimation method are based on collision slots in terms slot efficiency by
over 10% and also in terms of identification time.
Keywords:
Anti-collision
Dynamic frame slotted
ALOHA
Electronic product code
Internet of things
Performance evaluation
Radio frequency identification
Tag grouping
This is an open access article under the CC BY-SA license.
Corresponding Author:
Nnamdi H. Umelo
Department of Computer and Communication Systems Engineering, Faculty of Engineering, University
Putra Malaysia
Serdang 43400, Selangor, Malaysia
Email: gs54341@student.upm.edu.my
1. INTRODUCTION
Internet of things (IoT) has been defined in different ways by different stakeholders. Obviously, IoT
is different things to different people. Whereas some professionals tend to define IoT based on how they will
contribute to it, users on the other hand define IoT based on how they may benefit from it. But one consensus
definition of IoT is that it is that evolutionary trend of the internet where objects and just anything can
exchange data about itself and its surroundings. Evans [1] of Cisco Internet Business Security Group opined
that IoT is a state of Internet when objects gets connected and exchange information more that humans. He
argued that since the population of things already connected to the internet has surpassed human population
in 2008, we are already experiencing IoT. He made a prediction that in 2020 as the population of the world is
expected to increase to 7.6 billion, objects connected to the internet are expected to skyrocket to about
50 billion.

Int J Elec & Comp Eng ISSN: 2088-8708 
Grouping based radio frequency identification anti-collision protocols for dense … (Nnamdi H. Umelo)
5849
Radio frequency identification (RFID) is seen as an inevitable enabling technology for the
attainment of IoT’s vision of automatic identification and data communication of just everything [2]. RFID is
used for diverse applications like in health monitoring [3], shopping mall [4] and in the passports of USA,
UK and Malaysia. RFID has numerous advantages over other identification procedures like barcodes,
biometrics, and magnetic stripes. in that it can identify things wirelessly, simultaneously and within a long
range.
There are three types of RFID based on energy source of their tags namely: active RFID,
semi-passive RFID and passive RFID. Whereas both active RFID and semi-passive RFID have batteries
installed on each tag, only the active RFID tag can generate its own wave. As shown in Table 1, semi-passive
RFID though equipped with batteries, still rely on the backscattering of wave from reader query just like the
passive RFIDs. In Table 1, we show the characteristics of these three types of RFID. Passive ultra-high
frequency (UHF) RFID is the RFID for IoT as they are cheap and require no batteries [5]. Hence, they can
easily be deployed on a very large scale for object identification and sensing. Passive RFIDs are often used
interchangeably as RFID itself. Therefore, in this paper we shall be referring passive RFID as RFID.
Table 1. Categorization of RFID based on energy source of their tags
RFID Type Passive RFID Semi-passive RFID Active RFID
Battery presence NO YES YES
Communication principle Backscattering Backscattering Generates own wave
Communication range 10 cm to 7 m Less than 80 m Over 100 m
Cost ($) [5] Less than 0.5 1 to 2 Over 10
Operating frequency LF (125KHz), HF (13.56 MHz), UHF,
Microwave
UHF (860-960 MHz) UHF, Microwave
(2.45 GHz)
Application areas Access control, security, manufacturing
process, animal identification, smart cards,
supply chain
Library books tracking, animal
monitoring and identification,
vehicle immobilizers.
Aerospace, Railroad car
tracking, container
tracking
RFID suffers three types of collision namely: reader-to-reader collision, reader-to-tag collision, and
lastly tag-to-tag type of collision. Tag-to-tag collision is referred to as tag collision problem (TCP) of RFID
networks [6]. As depicted in Figure 1, TCP is a situation where tags’ signal (T1, T2) collides as they try to
respond to reader through backscattering. This is possible because tags use same communication channel to
reply to reader queries.
Figure 1. Tag-to-tag collision
In 2005, electronic product code (EPC global) standardized EPC class 1 generation 2 (EPC C1G2)
[7], [8] as a DFSA-based media access control (MAC) protocol that addresses TCP for passive ultrahigh
frequency (UHF) RFID and for a few tag network. However, as the number of tags increases, the efficiency
of EPC C1G2 deteriorates drastically [9]–[11]. Hence, the need to enhance DFSA for high tag density
applications like IoT. The focus of paper is on enhancing EPC C1G2 to support IoT application.
There have been several efforts to addressing TCP. In literature, approaches by different researchers
is categorized into two namely: ALOHA-based approach and tree-based approach [12]–[15]. Tree-based
protocols also referred to as deterministic based approach are divided into two, namely: Binary tree protocols
(binary splitting) and query tree protocols (QT). Binary tree protocols use random binary number to resolve
collision among tags while in QT protocols, the reader uses prefixes in tag IDs to resolve collision. The merit
of tree-based protocols is the fact that there is no tag starvation as all tags will eventually be identified albeit,
they are computationally complex as they consume more bits and time for tag identification [16]. On the

 ISSN: 2088-8708
Int J Elec & Comp Eng, Vol. 12, No. 6, December 2022: 5848-5860
5850
other hand dynamic frame slotted ALOHA (DFSA-based) (ALOHA-based) protocols are easy to implement
[6], ensures fairness in the random access of the shared channel and are promising for high density
application of RFID if enhanced and made robust to quickly mitigate collision when they occur.
DFSA is the De Facto algorithm adopted by EPC C1G2 protocol. however, the traditional DFSA
still has the challenge of poor tag number estimation and bogus frame size adjustment algorithm which
causes either too many idle or collision slots and consequently, affects channel utilization and efficiency of
the RFID system. IoT deployment of RFID demands accurate unidentified tag number estimation that would
effectively support dense deployment of tags. In the current tag number estimation algorithm, the RFID
reader does not have prior knowledge of tag number before the actual process of identification and relies on
querying the tags within its read range and estimating tag number using the number of collision slots from the
query result. The current estimation algorithm is based on an assumption that a collision means two tags have
collided [14]. The equation (1) is the lower bound of collision slot-based tag number estimation which
assumes that every collision slot means two tags have collided.
Over the years, researchers have tried to improve on this by proposing probability models that tries
to determine the actual value of the collision. However, the works of [17], [18] show that collision slots could
occur in an RFID system due to an error prone channel or interference from other signals. Therefore,
determining the value of collision slots alone is not enough to guarantee the accuracy of the tag estimation
algorithm and protocols developed for IoT with such assumption for collision slots could cause a total system
breakdown.
𝑛
̂ = 𝑆𝑠 + 2𝑆𝑐 (1)
Where 𝑛
̂ is the tag number estimate per query cycle, 𝑆𝑠 is the number of tags in the success slot per query
cycle and 𝑆𝑐 is the number of tags in the collision slot per query cycle.
Furthermore, in DFSA a frame is a collection of time slots, and each RFID tag can only be identified
by the reader using a time slot. In frame slotted ALOHA (FSA) and DFSA, the reader queries the tags using
frame per cycle unlike in slotted ALOHA (SA). In DFSA each query cycle tries to automatically decrease or
increase the number of time slot in a frame (frame size). Hence, the dynamism in DFSA. It follows that
efficiency of the RFID systems is a function of how close the frame size is to the estimated number of
unidentified tags. Traditionally, slot efficiency 𝑆𝐸 expressed in (2) is the most popular metric in evaluating
the performance of an RFID anti-collision protocol [9]. Hence, a good protocol is one that ensures the
efficient utilization of time slots (communication channel).
𝑆𝐸 =
𝑆𝑠+𝑆𝑐
𝑆𝑠+𝑆𝑖+𝑆𝑐
(2)
Where 𝑆𝑠 is number of successful slots, 𝑆𝑖 is number of ideal slots, and 𝑆𝑐 is number of collision slots.
On the other hand, Tag grouping is becoming a popular technique in RFID anti-collision research
[9]. The advantage of grouping can be summarized in that the number of tags is small (per group), and the
frame size is equal to the number of tags. This is to say that whereas DFSA gives fair share to all tags in
accessing the shared channel while responding to reader queries, grouping based schemes would only present
a fewer number of tags (a group or cluster) to the channel at a time even when the RFID tags are so densely
populated (IoT scenario).
2. GROUPING METHODS FOR RFID ANTI-COLLISION PROTOCOL
The underlining principle behind grouping is the fact that existing protocol EPC C1G2 can
effectively address collision in a few tags’ density. Theoretically and in practice, this principle ensures that
unlike FSA, DFSA can surpass the optimal throughput of about 36.8% [19]. In this section, we first present
an analysis of DFSA for RFID then we discussed each of the selected grouping-based algorithms.
2.1. Analysis of DFSA for RFID
Dynamic frame slotted ALOHA (DFSA) algorithm is a more complex algorithm that equips RFID
readers with the ability to dynamically adjust their frame sizes based on the intelligence or knowledge they
gather from the RFID system. DFSA addresses the limitation of FSA in terms of readers’ fixed (static) frame
sizes that either results to high collision rate or wastage of the communication channel depending on the tag
density of the RFID environment [20]. Figure 2 describes how DFSA works. From the diagram there are four
RFID tags that are to be identified by the reader. During the first read cycle (using frame 1), the reader
queries tags and tag 1 and tag 4 randomly chose slot 2 and slot 4 respectively and got successfully read by the

5851
reader while tag 2 and tag 3 randomly chose slot 3 at the same time causing a collision. Apart from the
collision, slot 1 and slot 5 were idle causing wastage of bandwidth. However, during read cycle 2 the reader
automatically reduced the frame size from 5 to 3 (frame 2) because there are only two remaining tags to be
read. However, there were still collision and idle slots. Note that slots 2 and 4 are referred to as success slots
S, slot 3 is referred to as collision slot C, while slots 1 and 5 are referred to as idle slots I.
The basic factor in DFSA is the ability of the reader to accurately estimate the number of
unidentified tags after each query cycle and have a precise frame size adjustment algorithm [21]. Existing
works in literature use collision slots as the basis for tag number estimation. The current RFID MAC
protocol: EPC class 1 generation 2 (EPC C1G2) is DFSA based. However, its efficiency is not guaranteed
with high tag density application like IoT. This paper evaluates grouping-based RFID algorithms and
compares them to traditional DFSA. Therefore, before discussing the selected grouping algorithms, we
present an analysis of DFSA algorithm for RFID. As stated earlier the only factor that determines what the
frame size should be per query cycle is the estimated number of unidentified tags.
Figure 2. DFSA anti-collision algorithm for RFID
In analyzing what the estimated tag number should be in relation to the frame size we assume that
the probabilities that tags choose slots are equal (fair access to the shared channel). Therefore, where 𝑛 is the
estimated number of tags and 𝐿 is the frame size and using the binomial distribution, the probability that
there are 𝑟 tags in a slot is given by:
𝑎𝑟 = (𝑛
𝑟
) (
1
𝐿
)
𝑟
(1 −
1
𝐿
)
𝑛−𝑟
(3)
Hence, the probability that there is only one tag in a slot is given by:
𝑎1 = 𝑛 (
1
𝐿
)
𝑟
(1 −
1
𝐿
)
𝑛−1
(4)
While the probability that there is no tag in a slot is given by:
𝑎0 = (1 −
1
𝐿
)
𝑛
(5)

 ISSN: 2088-8708
5852
Consequently, given the frame size 𝐿, in a query cycle the number of success slot are:
𝑆𝑠 = 𝐿 ∗ 𝑎1 = 𝑛 (1 −
1
𝐿
)
𝑛−1
(6)
While the number of idle slots is:
𝑆𝑖 = 𝐿 ∗ 𝑎0 = 𝐿 (1 −
1
𝐿
)
𝑛
(7)
Therefore, the number of collision slots is:
𝑆𝑐 = 𝐿 − 𝑆𝑠 − 𝑆𝑖 = 𝐿 − 𝑛 (1 −
1
𝐿
)
𝑛−1
− 𝐿 (1 −
1
𝐿
)
𝑛
(8)
Efficiency of the RFID system can be calculated as (9):
𝜂 =
𝑆𝑠
𝐿
(9)
To maximize efficiency, we calculate:
𝑑𝑆𝑠
𝑑𝐿
= (1 −
1
𝐿
)
𝑛−1
+ 𝑛 (1 −
1
𝐿
)
𝑛−1
ln (1 −
1
𝐿
) = (1 −
1
𝐿
)
𝑛−1
[1 + 𝑛 ln (1 −
1
𝐿
)] = 0 (10)
When 𝑛 is known:
𝐿 =
1
1−𝑒
1
𝑛
=
𝑒
1
𝑛
𝑒
1
𝑛−1
When the value of 𝑛 is very large:
𝐿 =
1+
1
𝑛
1+
1
𝑛
−1
= 𝑛 + 1 (11)
Therefore, it is proven in DFSA that the closer the frame size 𝐿 is to number of unidentified tags 𝑛,
the higher the efficiency of the RFID system. Moreover, in Figure 3, we use simulation to validate this
theoretical model in (11) with a tag density of 300. From Figure 3, it can be seen that for different frame sizes
𝐿 (16, 32, 64, 128, 256), efficiency is maximum when tag density 𝑛 is almost equal to 𝐿.
Figure 3. System Efficiency as a function of tag number in DFSA

5853
2.2. Fuzzy C-means grouping algorithm
Fuzzy C-means grouping algorithm (FCMBG) is a DFSA based anti-collision scheme that uses
fuzzy C-means machine learning technique together with Mahalanobis distance measure to group RFID tags
using their EPC codes. The algorithm is set up in such a way that the reader uses the grouping condition of
tags to predict frame size that would cause minimal collision. In the fuzzy C-means algorithm [19], RFID
tags are assumed to be 𝑛 vectors that are firstly grouped (𝑋 = {𝑥1𝑥2 … 𝑥𝑛}) into c fuzzy sets (𝑆1𝑆2 … 𝑆𝑛)
with c clustering centers as set 𝑣 = {𝑣1𝑣2 … 𝑣𝑛}. Then the cluster center of each group is computed so that
the value of the objective function is minimized. The fuzzy membership degree 𝑢𝑖𝑘 can be computed for each
tag using (12) while the centroid is computed using (13). The objective function is stated in (14):
𝑢𝑖𝑘 =
1
∑ (
𝑑𝑖𝑘
𝑑𝑗𝑘
)
2
𝑚−1
𝑐
𝑗=1
(12)
𝑣𝑖 =
∑ (𝑢𝑖𝑘)𝑚
𝑛
𝑘=1 𝑥𝑘
∑ (𝑢𝑖𝑘)𝑚
𝑛
𝑘=1
(13)
𝐿 ≈
1+
1
𝑛
1+
1
𝑛−1
= 𝑛 + 1, 𝑛 ≫ 1 (14)
Here is the procedure for grouping in FCMB grouping algorithm:
- Step 1: initialize the cluster center (centroid) 𝑣, number of groups 𝑐, fuzzy weighted index 𝑚 and range of
frame size [𝐿𝑚𝑖𝑛𝐿𝑚𝑎𝑥];
- Step 2: compute the fuzzy membership degree matrix 𝑢 using (12)
𝑢𝑖𝑗 =
{
[∑
‖𝑥𝑖−𝑣𝑗‖
2
𝑚−1
‖𝑥𝑖−𝑣𝑘‖
2
𝑚−1
𝑐
𝑘=1 ]−1 ‖𝑥𝑖 − 𝑣𝑘‖ ≠ 0
1 ‖𝑥𝑖 − 𝑣𝑘‖ = 0 = 𝑗
0 ‖𝑥𝑖 − 𝑣𝑘‖ = 0 ≠ 𝑗
(15)
where 𝑢𝑖𝑗 is the fuzzy membership degree given that 𝑥𝑖 belong to group 𝑗, and 𝑣𝑗 is the centroid of
group 𝑗.
- Step 3: update each centroid 𝑣 according to (16)
𝑣𝑗 =
∑ 𝑢𝑖𝑗
𝑚
𝑛
𝑖=1 𝑥𝑖
∑ 𝑢𝑖𝑗
𝑚
𝑛
𝑖=1
(16)
where 𝑛 is the number of primary data vector (tags).
- Step 4: using (17) the target number of groups is computed and if the minimum condition of the objective
function is not met go to step 2 else continue.
𝑐 = {
𝑐 − 1 𝑙𝑖 < 𝐿𝑚𝑖𝑛
𝑐 + 1 𝑙𝑖 > 𝐿𝑚𝑎𝑥
𝑐 𝐿𝑚𝑖𝑛 ≤ 𝑙𝑖 ≤ 𝐿𝑚𝑎𝑥
(17)
- Step 5: compute the number of tags for each group. If it falls within the range 𝑙𝑖 ∉ [𝐿𝑚𝑖𝑛 𝐿𝑚𝑎𝑥], update
the centroid and go to step 1, else STOP.
2.3. Group-based binary splitting algorithm
To ensure good coverage of the different approaches to RFID anti-collision in our evaluation of
grouping-based algorithms, this paper also selected a binary splitting based algorithm which is tree-based
grouping algorithm by [9] called group-based binary splitting algorithm (GBSA). In GBSA, tags are divided
into multiple subsets using a tag number estimation 𝑛 and an optimal grouping factor 𝑓𝑜𝑝𝑡 then identified
using a modified binary splitting (MBS) approach. The algorithm which is described in the block diagram in
Figure 4 starts with the reader performing the first query cycle by showing the initial frame size. After the
tags responds to the first query, the reader computes the 𝑛 𝑎𝑛𝑑 𝑓𝑜𝑝𝑡 using the statistics of the success,
collision, and idle slots. Consequently, the tags within the read range of the reader are grouped into
𝑁 = 𝑛. 𝑓𝑜𝑝𝑡 groups. Afterwards, the modified binary splitting is used for tag identification in each group. In

 ISSN: 2088-8708
5854
their modified binary splitting strategy, GBSA deployed a 1-bit random binary generator 𝑅 to pre-split
contending tags. Since the time used for collision arbitration is eliminated, hence, efficiency and throughput
are enhanced. For their tag number estimation function, GBSA adopted [22] and it is based on the result of
the first sub-frame. They considered 𝑛 tags are assigned 𝐿 slots, after each query cycle, the probabilities that
a slot is success slot 𝑆𝑠, or collision slot 𝑆𝑐or idle slot 𝑆𝑖 is given by:
𝑃(𝐿𝑠𝑢𝑏, 𝑆𝑠𝑆𝑐𝑆𝑖) = (
𝐿𝑠𝑢𝑏!
𝑆𝑠!𝑆𝑐!𝑆𝑖!
) × 𝑃0(𝑆𝑖) × 𝑃1(𝑆𝑠|𝑆𝑖) × 𝑃2(𝑆𝑐|𝑆𝑖, 𝑆𝑠) (18)
where 𝑃1, 𝑃2 and 𝑃0 are probabilities of success slot, collision slot and idle slot. The estimated number of tags
in a given sub-frame is gotten when the probability of (18) is maximum. This means that 𝑛
̂𝑠𝑢𝑏 = 𝑛 when
𝑃(𝐿𝑠𝑢𝑏, 𝑆𝑠𝑆𝑐𝑆𝑖) is maximum. Therefore, for the full frame 𝐿, the estimated tag number is computed:
𝑛
̂𝑒𝑠𝑡 = 𝑛
̂𝑠𝑢𝑏 ×
𝐿
𝐿𝑠𝑢𝑏
(19)
Figure 4. Function block diagram for GBSA
2.4. Grouping-based DFSA with fine groups (GB-DFSA)
One of the first works on grouping based of RFID anti-collision algorithms was proposed by [23].
Their algorithm has two stages. Firstly, the grouping phase and then the identification phase. Grouping is
achieved in the first phase using the fact that RFID systems attain maximum throughput when the number of
unidentified tags equals the frame size. Evenly, maximum throughput is attained when the tags are small, and
the frame size is equally small. However, as the number of tags decreases, throughput increases hence, they
performed the partitioning of tags into smaller groups and then identify with DFSA using Schoute’s tag
number estimation method [24]. In GB-DFSA shown in Table 2, the RFID reader first estimates number of
groups 𝑠, by performing few queries then it enters into the identification phase where the reader arranges the
tags into 𝑠 groups then use DFSA to read the tags. The algorithm has six variables: 𝑓 represents the frame
size, 𝑠 represents number of groups, 𝑝𝑡ℎ is for threshold, 𝑖 is number of queries in the grouping phase, 𝑚 is
fraction factor. 𝑡𝑎𝑔𝑠𝑔 represents tags within a group.

5855
Table 2. GB-DFSA grouping algorithm
GB-DFSA Grouping Algorithm
GB-DFSA(tags) {
𝑓=4; //Frame size
𝑠=1; //the number of groups
𝑝𝑡ℎ=8; //Threshold
𝑖=1; //the number of iterations
𝑚=4;
𝑡𝑎𝑔𝑠𝑔=𝑡𝑎𝑔𝑠;
//grouping phase
do{
𝑛𝑐=FSA(𝑡𝑎𝑔𝑠𝑔, 𝑓);
If (2.39*𝑛𝑐>𝑝𝑡ℎ);
i=i+1;
𝑠=𝑚𝑖−1
;
[𝑔𝑟𝑜𝑢𝑝1, … , 𝑔𝑟𝑜𝑢𝑝𝑠]=grouping (𝑡𝑎𝑔𝑠, 𝑠);
//randomly choose one of the groups}
𝑡𝑎𝑔𝑠𝑔=random ([𝑔𝑟𝑜𝑢𝑝1, … , 𝑔𝑟𝑜𝑢𝑝𝑚
𝑖−1
]);
end if
}until (2.39*𝑛𝑐 <= 𝑝𝑡ℎ)
//identification phase
𝑓 = 𝑝𝑡ℎ;
𝑔𝑟𝑜𝑢𝑝 = 𝑔𝑟𝑜𝑢𝑝1;
𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑡𝑎𝑔𝑠 =DFSA(𝑔𝑟𝑜𝑢𝑝, 𝑓);
for i=2 to 𝑠 do{
𝑓 = 𝑛𝑢𝑚𝑏𝑒𝑟(𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑡𝑎𝑔𝑠)/(𝑖 − 1);
𝑔𝑟𝑜𝑢𝑝 = 𝑔𝑟𝑜𝑢𝑝1;
𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑡𝑎𝑔𝑠
= 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑡𝑎𝑔𝑠 ᴗ DFSA(𝑔𝑟𝑜𝑢𝑝, 𝑓);
}
return 𝑖𝑑𝑒𝑛𝑡𝑖𝑓𝑖𝑒𝑑 𝑡𝑎𝑔𝑠
}
3. PERFORMANCE EVALUATION/RESULT AND DISCUSSION
In this section, we present an evaluation of the performance of the three selected grouping-based
algorithms for RFID anti-collision together with the traditional DFSA. For fairness and better coverage, the
selection considered DFSA based RFID anti-collision algorithms in conformity with EPC C1G2 and one
tree-based grouping algorithm. All simulations were performed in MATLAB software (8.5.0, MathWorks,
Natick Massachusetts, USA). Different scenarios of tag density and frame sizes were used to generate results
in an extensive Monte Carlo simulation. We assume the communication channel is devoid of any capture
effect and noise. In RFID anti-collision research, there are two main metrics that can be used to derive the
appropriate frame size of the RFID system namely throughput rate and channel utilization factor
(slot efficiency) [25]. We also compared the algorithms performances in terms of time consumption which
[26] defined as total number of slots used by an algorithm to read all tags successfully.
3.1. Simulation parameters and data
Electronic product code (EPC) is a unique identifier of all RFID tags. Just like barcodes, they are
used to identify any physical object in the world uniquely. This attribute is one of the reasons RFID is
imperative to attaining IoT, which is to enable just anything have connectivity. Tag EPC is 96-bits and is
accessible in the cloud via EPC information service (EPCIS). As shown in Table 3, EPC is long and in
conformity with EPC C1G2, this paper uses only 16-bits out of the 96-bits for our evaluation. As shown in
Table 3, the EPC code has four parts. The first 8 bits specifies the version of RFID tag while in the domain
management, the manufacturer identifies the tag uniquely from the organizational side. The object class of
24-bit is used by the EPC management to identify the type of item while the 36-bit serial number is unique
within each object class. This unique serial number forms the basis for our grouping in this paper as 16-bit
binary data are generated randomly in MATLAB to simulate the tags’ RN16. In our study, the 16-bit data ID
is used in the clustering (grouping) of tags for each algorithm-traditional DFSA, GBSA, GB-DFSA and
FCMBG.
Table 3. RFID tag coding
EPC 96-bit ID Version no. Domain Management Object Class Serial no.
No. of bits 8 28 24 36

 ISSN: 2088-8708
5856
The simulation was under a single reader and multiple tags scenario with assumed error-free
channel. Also, part of the assumption was that no tags leave or enter the RFID read range. We set up a high
tag density RFID scenario with one reader. The tags ranges from 100 to 1,000 same as in [9]. The parameters
used in the simulation are in conformity with EPC C1G2 and are listed in Table 4. So, at the time of
simulation, 1,000 groups of 16-bit binary data were generated at random with the MATLAB software to
simulate the EPC code of RFID tags. Noteworthy is the fact that like in EPC C1G2 specifications, in our
evaluation, the different grouping algorithms were implemented using the different correct time slots value
for the success slot, collision slot and idle slots T1, T2 and T3, respectively.
Table 4. Simulation parameters
Parameters Value Parameters Value
Reader-to-tag data-0 1 Tari RTcal 37.5 µs
Reader-to-tag data-1 2 Tari TRcal 50 µs
Reader-to-tag rate 80 Kbps T1 62.5 µs
Tag-to-reader rate 160 Kbps T2 62.5 µs
Tpri 6.25 µs T3 100 µs
Tari 12.5 µs Probe 4 bits
feedback 3 bits RN16 16 bits
Query 22 bits ID 96 bits
Query Adj 9 bits Ack 18 bits
R-T Preamble 112.5 µs QueryRep 4 bits
T-R Preamble 37.5 µs Framesync 62.5 µs
3.2. Results and discussion
In the experiment we performed the effect of tag density on throughput rate of the RFID system
using the throughput rate formula in (20):
𝑇ℎ𝑟 =
𝑆𝑠.𝑇𝑠
𝑆𝑖.𝑇𝑖+𝑆𝑠.𝑇𝑠+𝑆𝑐.𝑇𝑐
(20)
where 𝑆𝑠 𝑎𝑛𝑑 𝑇𝑠 represent the value and time of success slots per query cycle, 𝑆𝑖 𝑎𝑛𝑑 𝑇𝑖 represent the value
and time of idle slots per query cycle and lastly 𝑆𝑐 𝑎𝑛𝑑 𝑇𝑐 represent the value and time of collisions slots per
query cycle.
As can be seen in Figure 5, different frame sizes 𝐿, were used to evaluate the throughput rate result
of the four algorithms. In Figure 5(a), given a frame size of 64, FCMBG shows edge over the other three
algorithms recording a throughput rate of over 55% which is far better than traditional DFSA which gives a
result of about 35% throughput while GBSA followed FCMBG with about 50% throughput rate. It can also
be seen in Figure 5(b) that GBSA shows more stability in throughput rate with very high tag density of 800
and above than FCMBG that shows decline in throughput rate. This can be attributed to tag starvation
problem of DFSA based protocols. A situation where some tags might never be read by the reader due to
very high population of tags unlike tree based protocols [27]. Consequently, this study discovers that the
sharp decline occurred only when frame size was minimal 𝐿=64. This means that whereas FCMBG gives an
improved throughput than traditional DFSA with increasing tag density, FCMBG works best when the tag
density is high and when the frame size is as large as the number of tags as seen in Figures 5(c) and 5(d).
This finding is in agreement with the theoretically analysis in (11).
Furthermore, in Figures 5(b), 5(c) and 5(d), given frame sizes 𝐿=128, 𝐿=256 and 𝐿=512
respectively, FCMBG shows highest throughput rate among the four algorithms while the other two
grouping-based algorithms; GB-DFSA and GBSA closely followed. DFSA throughput rate crashed with
increasing tag density. This huge difference in throughput rate between DFSA and the other three algorithms
is because unlike DFSA, tag grouping in the other algorithms is a proactive means of minimizing collision
even before the tags randomly respond to reader queries. Hence, from our study we deduce that using fuzzy
C-means algorithm to group or cluster RFID tags before running traditional DFSA yields a better throughput
rate when the number of RFID tags to be identified by the RFID reader are of high density.
Slot efficiency also referred to as channel utilization factor is the ratio of not empty slots and total
timeslot (frame). Slot efficiency expressed in (2) is a measure of whether a given protocol wastes
communication channel or not in the identification process. In DFSA based RFID anti-collision, the shared
channel is divided into time slots and a tag can only reply to reader queries using a slot. The more slots a
particular protocol uses to identify all of a given set of tags the lesser its channel efficiency [9], [12].
Therefore, slot efficiency is a very important metric in evaluating the performance of any RFID anti-collision
protocol. In our simulation result shown in Figures 6(a) to 6(d) unsurprisingly, GBSA has perfect slot

5857
efficiency. This is because tree-based protocols have the advantage of utilizing the channel well albeit, with a
consequential trade off which is too much time delay as seen in Figure 7. This delay is due to fact that
tree-based protocols are deterministic and cause too much computational space to implement. Hence, the
choice of DFSA in EPC C1G2 protocol due to limited computational space of passive tags. However,
FCMBG performed better than the other two DFSA based algorithms in terms of channel utilization.
(a) (b)
(c) (d)
Figure 5. System throughput rate comparison for the different algorithm using (a) frame size of 64, (b) frame
size of 128, (c) frame size of 256, and (d) frame size of 512
(a) (b)
(c) (d)
Figure 6. Slot efficiency (channel utilization factor) for the different algorithms using (a) frame size of 64,
(b) frame size of 128, (c) frame size of 256, and (d) frame size of 512

 ISSN: 2088-8708
5858
In comparing identification time used by each algorithm to identify all the tags, there is
inconsistencies among different authors in literature. For instance [19], [20] in presenting their results define
identification time as the total time slot number used in identifying all the tags while [28] considered both
total time slot number and total query rounds. However, for the purpose of this evaluation we define
Identification time as total number of timeslots utilized by each algorithm to identify all the tags. This is in
accordance with [19] because the work is part of our performance comparison. Going by the result presented
in Figure 7 and as expected GBSA which uses modified binary splitting consumes the highest time to
identify all tags.
Figure 7. Identification time for the different algorithms
However, it was surprising to observe that even when the tag density was as low as 100, GBSA
shows a result of over 200% more time to identify all 100 tags. This is associated with the recursive nature of
breaking collision slots into ones and zeros continuously to form a tree-like structure of tree-based algorithms
and binary splitting in particular. Based on this observation, we infer that for IoT application of RFID where
time and computational space is a challenge, binary splitting algorithms might not be suited.
On the other hand, FCMBG which uses fuzzy C-means machine learning to group tags performed
best in terms of identification time when the tag number is very dense (over 500). Initially when the tag
density was less than 300, the traditional DFSA, GB-DFSA and FCMBG all show low identification time.
But as the tag density increases to 500 and to 1,000, FCMBG shows much lower time than all the other three
algorithms. This could be attributed to the fact that unlike the traditional DFSA, FCMBG uses the uniqueness
of tags EPC (RN16 code) to group tags effectively, making it less probable for tags to randomly select same
time slot (collision). Consequently, minimal collision means minimal retransmission by the reader and lesser
time used in identifying all the tags within the read range of the reader.
4. CONCLUSION
Firstly, this paper validates through simulation the assertion that the efficiency of the RFID system
is maximum when the number of tags is almost equal to the frame size the reader uses to query the tags. The
paper also demonstrates the edge grouping-based RFID anti-collision algorithms have over traditional RFID
anti-collision algorithm when the number of tags to be identified are large like in IoT. The paper shows that
when the unique serial number of RFID tags are used to classify or group tags like in FCMBG protocol,
collision is further mitigated while system throughput and time delay are enhanced. Because RFID tags have
unique EPC codes, binary 16-bit codes were generated in MATLAB to model the tags, while different
grouping algorithms (FCMBG, GB-DFSA, GBSA) were implemented in MATLAB including traditional
DFSA algorithm. Results show that all the three selected grouping-based RFID anti-collision algorithms gave
better performances than traditional DFSA when tag density is high in terms of throughput rate, channel
utilization while FCMBG and GB-DFSA performed better in terms of identification time. The study also
demonstrates that machine learning could be the most efficient technique in performing RFID tag grouping

5859
since FCMBG gave the best result among the selected algorithms. Therefore, grouping based RFID
anti-collision algorithms are promising and could be the sine qua non for RFID MAC protocol development
for IoT application.
REFERENCES
[1] D. Evans, “How the next evolution of the internet is changing everything,” Cisco Internet Business Solutions Group (IBSG), no.
April, 2011.
[2] S. M. Florence, M. Uma, C. Fancy, and G. Saranya, “A study of remotely booking slot for vehicle using internet of things,”
International Journal of Electrical and Computer Engineering (IJECE), vol. 10, no. 5, pp. 5392–5399, Oct. 2020, doi:
10.11591/ijece.v10i5.pp5392-5399.
[3] H. A. Khan, R. Abdulla, S. K. Selvaperumal, and A. Bathich, “IoT based on secure personal healthcare using RFID technology
and steganography,” International Journal of Electrical and Computer Engineering (IJECE), vol. 11, no. 4, pp. 3300–3309, Aug.
2021, doi: 10.11591/ijece.v11i4.pp3300-3309.
[4] A. Abdulkareem, C. Awosope, and A. Tope-Ojo, “Development and implementation of a miniature RFID system in a shopping
mall environment,” International Journal of Electrical and Computer Engineering (IJECE), vol. 9, no. 2, pp. 1374–1378, Apr.
2019, doi: 10.11591/ijece.v9i2.pp1374-1378.
[5] S. Hodges and M. Harrison, “Demystifying RFID: principles and practicalities, auto-ID centre,” Institute for Manufacturing,
University of Cambridge, pp. 1–23, 2004.
[6] X. Wang, “A fast hybrid strategy-based RFID tag identification protocol,” Wireless Communications and Mobile Computing,
vol. 2021, pp. 1–7, Feb. 2021, doi: 10.1155/2021/6646812.
[7] G. EPCglobal, “Epc radio-frequency identity protocols generation-2 uhf rfid; specification for RFID air interface protocol for
communications at 860 mhz-960 mhz,” EPCglobal Inc., November, vol. 1, no. 3, 2013.
[8] H.-Y. Chien and C.-H. Chen, “Mutual authentication protocol for RFID conforming to EPC Class 1 Generation 2 standards,”
Computer Standards and Interfaces, vol. 29, no. 2, pp. 254–259, Feb. 2007, doi: 10.1016/j.csi.2006.04.004.
[9] J. Su, Z. Sheng, A. X. Liu, Y. Han, and Y. Chen, “A group-based binary splitting algorithm for UHF RFID anti-collision
systems,” IEEE Transactions on Communications, vol. 68, no. 2, pp. 998–1012, Feb. 2020, doi: 10.1109/TCOMM.2019.2952126.
[10] Z. Jiang, B. Li, M. Yang, and Z. Yan, “LC-DFSA: low complexity dynamic frame slotted aloha anti-collision algorithm for RFID
system,” Sensors, vol. 20, no. 1, Dec. 2019, doi: 10.3390/s20010228.
[11] W. H. Zhou, N. De Jiang, and C. C. Yan, “Research on anti-collision algorithm of RFID tags in logistics system,” Procedia
Computer Science, vol. 154, pp. 460–467, 2019, doi: 10.1016/j.procs.2019.06.065.
[12] J. Su, Z. Sheng, L. Xie, G. Li, and A. X. Liu, “Fast splitting-based tag identification algorithm for anti-collision in UHF RFID
system,” IEEE Transactions on Communications, vol. 67, no. 3, pp. 2527–2538, Mar. 2019, doi:
10.1109/TCOMM.2018.2884001.
[13] X. Jia, M. Bolic, Y. Feng, and Y. Gu, “An efficient dynamic anti-collision protocol for mobile RFID tags identification,” IEEE
Communications Letters, vol. 23, no. 4, pp. 620–623, Apr. 2019, doi: 10.1109/LCOMM.2019.2895819.
[14] Z. Huang et al., “A novel cross layer anti-collision algorithm for slotted ALOHA-based UHF RFID systems,” IEEE Access,
vol. 7, pp. 36207–36217, 2019, doi: 10.1109/ACCESS.2019.2900739.
[15] Y. Qiu, Y. Xu, L. Yang, J. Lu, and D. Li, “Improved anti-collision algorithm for the application on intelligent warehouse,”
Applied Sciences, vol. 9, no. 8, Apr. 2019, doi: 10.3390/app9081596.
[16] L. Zhao, L. Wang, and S. Du, “An efficient-algorithm for RFID tag anticollision,” Wireless Communications and Mobile
Computing, vol. 2021, pp. 1–7, May 2021, doi: 10.1155/2021/9967739.
[17] D. Benedetti, G. Maselli, C. Petrioli, and M. Piva, “The impact of external interference on RFID anti-collision protocols,” IEEE
Networking Letters, vol. 1, no. 2, pp. 76–79, Jun. 2019, doi: 10.1109/LNET.2019.2909814.
[18] Y.-H. Chen, R.-Z. Hung, L.-K. Chen, P.-T. Jan, and Y.-R. Su, “Channel-quality aware RFID tag identification algorithm to
accommodate the varying channel quality of IoT environment,” Applied Sciences, vol. 9, no. 2, Jan. 2019, doi:
10.3390/app9020321.
[19] B. Zhi, W. Sainan, and H. Yigang, “A novel anti-collision algorithm in RFID for internet of things,” IEEE Access, vol. 6,
pp. 45860–45874, 2018, doi: 10.1109/ACCESS.2018.2863565.
[20] H.-W. Wang, “Efficient DFSA algorithm in RFID systems for the internet of things,” Mobile Information Systems, vol. 2015,
pp. 1–10, 2015, doi: 10.1155/2015/942858.
[21] O. Bang, J. Hwan, D. Lee, and H. Lee, “Efficient novel anti-collision protocols for passive RFID tags,” AUTO-ID LABS White
Paper, 2009.
[22] Y. Chen, J. Su, and W. Yi, “An efficient and easy-to-implement tag identification algorithm for UHF RFID systems,” IEEE
Communications Letters, vol. 21, no. 7, pp. 1509–1512, Jul. 2017, doi: 10.1109/LCOMM.2017.2649490.
[23] C.-Y. Wang and C.-C. Lee, “A grouping-based dynamic framed slotted ALOHA anti-collision method with fine groups in RFID
systems,” in 2010 5th International Conference on Future Information Technology, 2010, pp. 1–5, doi:
10.1109/FUTURETECH.2010.5482699.
[24] F. Schoute, “Dynamic frame length ALOHA,” IEEE Transactions on Communications, vol. 31, no. 4, pp. 565–568, Apr. 1983,
doi: 10.1109/TCOM.1983.1095854.
[25] W. -T. Chen, “An accurate tag estimate method for improving the performance of an RFID anticollision algorithm based on
dynamic frame length ALOHA,” IEEE Transactions on Automation Science and Engineering, vol. 6, no. 1, pp. 9–15, Jan. 2009,
doi: 10.1109/TASE.2008.917093.
[26] H. Guo, C. He, N. Wang, and M. Bolic, “PSR: a novel high-efficiency and easy-to-implement parallel algorithm for anticollision
in RFID systems,” IEEE Transactions on Industrial Informatics, vol. 12, no. 3, pp. 1134–1145, Jun. 2016, doi:
10.1109/TII.2016.2554465.
[27] S.-C. Tsai, Y.-M. Hu, C.-H. Chai, and J.-S. Li, “Efficient tag reading protocol for large-scale RFID systems with pre-reading,”
Computer Communications, vol. 88, pp. 73–83, Aug. 2016, doi: 10.1016/j.comcom.2016.04.019.
[28] I. Onat and A. Miri, “A tag count estimation algorithm for dynamic framed ALOHA based RFID MAC protocols,” in 2011 IEEE
International Conference on Communications (ICC), Jun. 2011, pp. 1–5, doi: 10.1109/icc.2011.5962792.

 ISSN: 2088-8708
5860
BIOGRAPHIES OF AUTHORS
Nnamdi H. Umelo is a PhD candidate in the Department of Computer and
Communication Systems Engineering, (WiPNET Lab) UPM, Malaysia. He holds MSc from
the Department of Computer Science and Engineering, OAU Ile-Ife, Nigeria in 2012,
5962792and a B.Engr. Computer Engineering from ESUT, Enugu, Nigeria in 2006. His
research interest is in RFID for IoT, WSN, Wireless Networks, and Smart homes. He can be
contacted at email: umelonnamdi@gmail.com.
Nor K. Noordin has been a Professor in UPM since 2012 where she had earlier in
2006 gotten her PhD. In 1987, she got her B.Sc. in Telecommunications Engineering from the
University of Alabama USA, and afterwards a M.Sc. degree from UTM Malaysia. She is
currently the Dean Faculty of Engineering, UPM, Malaysia. She has over 200 journal and
conference papers and has attracted and conducted over 30 research projects. She can be
contacted at email: nknordin@upm.edu.my.
Mohd Fadlee A. Rasid is currently a Professor in the Department of Computer
and Communications Systems Engineering, UPM. He received a B.Sc. in Electrical
Engineering from the Purdue University USA and a PhD in Elect/Elect Engineering (Mobile
Communications) from Loughborough University UK. He is the director of WSN research
group which focuses on wireless medical sensors in healthcare. He has been a research
consultant to different projects including British council UKIERI project on wireless sensor
medical research. He can be contacted at email is: fadlee@upm.edu.my.
Tan K. Geok received the B.E., M.E., and PhD. degrees all in Electrical
Engineering from UTM, Malaysia, in 1995, 1997, and 2000, respectively. He has been Senior
R&D engineer in EPCOS Singapore in 2000. In 2001–2003, he joined DoCoMo Euro-Labs in
Munich, Germany. He is currently academic staff in MMU, Malaysia. His research interests
include radio propagation for outdoor and indoor, RFID, multi-user detection technique for
multicarrier technologies, and A-GPS. He can be contacted at email: kgtan@mmu.edu.my.
Fazirulhisyam Hashim received his M.Sc. degree from Universiti Sains Malaysia
and Ph.D. degree in telecommunication engineering from the University of Sydney, Australia.
He is currently an Associate Professor with the Department of Computer and Communication
Systems Engineering, Universiti Putra Malaysia. His research interests include heterogeneous
wireless communication systems and network security. He can be contacted at email:
fazirul@upm.edu.my.

Grouping based radio frequency identification anti-collision protocols for dense internet of things application

More Related Content

Similar to Grouping based radio frequency identification anti-collision protocols for dense internet of things application (20)

More from IJECEIAES (20)

Recently uploaded (20)

Grouping based radio frequency identification anti-collision protocols for dense internet of things application