Received 18 October 2023, accepted 13 November 2023, date of publication 20 November 2023,

date of current version 27 November 2023.

Digital Object Identifier 10.1109/ACCESS.2023.3334636

Anti-Collision Algorithm for Identification in

Precision Agriculture Applications
MAJID ALOTAIBI 1 , MOHSIN MURAD 1 , IMRAN ALI TASADDUQ1 , SHAKIR A. H. ALHUTHALI2 ,
FAISAL R. AL-OSAIMI 1 , FAHD ALDOSARI 1 ,
AND ALI MUHAMMAD RUSHDI 2 , (Life Senior Member, IEEE)
1 Department of Computer Engineering, Umm Al-Qura University, Makkah 21955, Saudi Arabia
2 Department of Electrical and Computer Engineering, King Abdulaziz University, Jeddah 21589, Saudi Arabia
Corresponding author: Mohsin Murad ([email protected])
This work was funded by the National Science, Technology and Innovation Plan (NSTIP), King Abdulaziz City for Science and
Technology (KACST), Kingdom of Saudi Arabia, grant number (14-ELE1448-10).

ABSTRACT Precision Agriculture (PA) techniques employing Internet of Things (IoT) can significantly
improve crop yields and enhance productivity. Radio frequency identification (RFID) based IoT systems
can enhance precision livestock farming by enabling livestock traceability and identification in large farms,
but congestion is a major challenge. In such conditions, livestock estimation using RFID is a complex task
because of high chances of tag collision due to significant increase in interference at the gates. Several
anti-collision algorithms have been explored in recent past to address this issue with the aid of a single
antenna. However, the throughput gains from a single antenna are limited particularly for dense scenarios
like gates at a small ruminant farm. Therefore, we propose a multi-antenna RFID reader system with a
dynamic frame slotted ALOHA (DFSA) anti-collision algorithm to address the tag collision problem by
exploiting the spatial diversity of the multiple inputs and multiple outputs (MIMO) backscattering channel.
The proposed technique is simulated in MATLAB for several tag population figures both with and without
interference actions of various intensities. Results have shown that significant throughput gains can be
achieved, in comparison with gains attained by anti-collision algorithms based on a single antenna. The
simulation testbed enables us to evaluate the proposed technique for multiple receive antenna configurations.

INDEX TERMS Livestock, farming, rfid, mimo, precision, agriculture, multi-antenna, anti-collision.

I. INTRODUCTION systems with signals from multiple transponders interfering

Precision Agriculture (PA) techniques are being widely with each other. As the farms scale and herd sizes increase,
adopted by farms around the world due to rising global pop- accurate reading with low errors is a significant challenge due
ulation, the need for higher crop yields as well as financial to high interference at the identification gates. This is more
incentives because of improved production [1]. PA techniques common in small ruminant farms [5] where animal density
based on the Internet of Things (IoT) can enhance production is very high and human staff is less. Through wireless signal
and significantly reduce human intervention [2]. RFID has transmission, an IoT system with radio frequency identifica-
widely been used for livestock traceability [3] and identifica- tion (RFID) enables people to remotely recognize objects [6],
tion applications [4] and modern farms now rely on tagging [7]. Access control, transportation, object detection, digital
their cattle for identification purposes. The common form identification and tracking are just a few of the areas in which
of tags used in ruminant farms are collar, ear tags, ruminal it has found use [8], [9], [10]. While such systems have reli-
bolus and under skin injectable tags. Simultaneous transpon- ably been used in a variety of access control applications such
der reading has always been a shortcoming of legacy RFID as public transportation, they are prone to the tag collision
problem when deployed in dense environments. Such as a
The associate editor coordinating the review of this manuscript and small ruminant farm with high herd density requiring a large
approving it for publication was Geng-Ming Jiang . number of tags.

2023 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.
VOLUME 11, 2023 For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ 130197
 M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

reader by adjusting impedance load of its associated antenna

in accordance with a specific code stored in the tag and
known as identity (ID) data. The required information will
be extracted by the reader through estimation and decoding
of the signal transmitted by the tag.
Several studies have been conducted to investigate
multiple-input multiple-output (MIMO) [12], [24] scheme for
various RFID readers. Significant increases in bit error rate
(BER) and range enhancement of reverse-link interrogation
(RIR) can be realized by using MIMO technology in the
backscatter channel. By improving throughput and read reli-
ability, we can use the designed readers in a variety of critical
applications such as large tag reading scenarios on the gates
of small ruminant farms, inventory management and crowd
estimation, among others. Estimation of tag quantities in bulk
reading scenario is directly proportional to the throughput of
FIGURE 1. A typical MIMO RFID reader system.
the system. Most of the estimation techniques, proposed for
RFID systems with a single antenna, typically underperform
in the above-mentioned critical situations. This is a result of
An RFID system consists mainly of RFID tags and readers high interference and low signal to noise ratio (SNR) due to
as well as middleware software that uses application pro- overlapping of tags’ signals.
gramming interfaces (APIs) [11], [12]. An antenna, or several Figure 1 depicts a block diagram of our proposed MIMO
antennae, and a tiny electrical circuit with a special ID make RFID system, wherein we consider an RFID system with
up the tag. There are several varieties of tags, including multiple receive antennas. A single antenna at the transmitter
passive and active ones [13]. A passive tag depends on the is considered, because the transmit antennas at the reader are
strength of the received signal from the transmitter module only serving the purpose of powering up the tags, by trans-
of the RFID reader and modulates it, whereas an active tag mitting similar unmodulated waves to the tag. Since the tag
has its own power source—usually a built-in battery—that circuits perform the coding and modulation operations, the
powers the tag’s circuit and enables radio frequency (RF) channel gains are low. For transmit antennas, the statistical
communication [14]. Passive RFID tags only rely on the properties of utilizing a single one or many of them at the
carrier wave’s energy from the reader because they lack an reader are the same when the data is complex Gaussian
internal power source. They are used in short-range applica- distributed and the transmitting energy is normalized [24].
tions because their range is little compared to the active tags. Multiple tags responding concurrently to the reader can
The sensitivity of the tag chip limits the range, which can undergo collisions and the situation deteriorates further in
be extended by using numerous antennae in an RFID reader large tag population environments [25]. Since passive RFID
[15]. Similarly, tag design can have a substantial impact on tags lack their own power source and transmitter, they are
sensitivity, and several efforts have been made to enhance this unable to interact with other tags that are already present
aspect of the system [15], [16], [17], [18]. in the surrounding region and are therefore unable to detect
Passive tags are an appealing alternative for applications these collisions [26]. When collision happens in a particular
where cost cutting is necessary and extended shelf life is time slot, the reader is unable to read the tags that collided,
desired [19]. Various channel measurements for RFID, like necessitating a time-consuming and energy-wasting repeti-
those carried out in [20] and [21], demonstrated it can be tion of the identification procedure. Anti-collision algorithms
categorized as a channel which is cascaded in nature having are used to address this issue by attempting to reduce the
forward as well as backscattering components, and both of likelihood that a collision would occur. The development of
these components can be represented using the Rayleigh anti-collision algorithms aims to accelerate the identification
distribution [22]. It is pertinent to mention that the cascaded method, lower the likelihood of collision, and thereby boost
channel degrades transmission since it fades faster than a the overall effectiveness of RFID performance. Algorithms
regular Rayleigh channel [23]. The reader-tag separation, tag based on ALOHA and Tree-based techniques are two major
placement with respect to the receiver, noise, interference, categories for anti-collision algorithms.
and multipath fading are some of the elements influenc- In this work, we focus on ALOHA based Dynamic Frame
ing the reader’s tag reading capacity. Interference can be Slotted ALOHA algorithm (DFSA). DFSA techniques [27],
caused by numerous tags emitting signals at the same time [28] of handling collisions in single antenna RFID systems
or by RFID readers nearby. In backscatter communication, have been around for quite some time [29] and they are able to
an RFID reader will initially broadcast a continuous radio significantly improve the throughput of the reader. In a DFSA
wave (CW) to activate the tag or multiple tags. This is fol- algorithm, a frame’s initial number of slots is set, and tags
lowed by the tag returning the received signal back to the are assigned at random. The reader then reads in the assigned

130198 VOLUME 11, 2023

M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

tolerance, that is overlooked in typical collision recovery

approaches, according to [31], can be utilized for signif-
icant gains in recovery. Backscatter link frequency (BLF)
is considered to be the same for colliding tags in typical
collision recovery techniques. However, because BLF might
vary depending on manufacturing procedures, tolerance must
be considered, but an accurate channel estimate is required.
Further gains can be made by artificially replicating BLF
tolerance. The approach involves correctly decoding a single
tag from a collection of many colliding tags in a single slot.
Each antenna’s power spectrum is computed and then it is
divided by BLF to get the energy spectrum.
FIGURE 2. The dynamic frame slotted ALOHA technique.
Collision happens when two or more tag signals are
received in the same time slot [32]. In the scenario where the
count of readers is smaller than the tag count, a communica-
tion model based on multiple labels can improve interrogation
time slots (Figure 2) and if no collision is detected, the reading
performance by employing non-sparse techniques [33]. Li et
process terminates, but if collisions are detected, the process
al. [34] investigate the scenario of several tags, which necessi-
is repeated by computing a different frame size considering
tates the incorporation of a multi antenna RFID model using a
the numbers of idle slots, collisions occurring and successful
MIMO scheme. An RFID model with M reader antennas, and
reads. The reader thus selects whether to increase, reduce,
N tags is assumed. A novel approach for recognizing multiple
or keep the previous frame length for the upcoming frame.
tags is described, which involves labelled collision-based
However, the frame length for the upcoming frame is
blind source signal separation followed by recovery. An anti-
reduced if the number of colliding slots is less than a certain
collision technique based on the Searching and Averaging
lower threshold. However, if the number of colliding slots is
Method in the Time Domain (SAMTD) is proposed for bulk
within acceptable limits (i.e., sandwiched between the two
tag reading. Furthermore, a two-step extraction strategy for
thresholds), the reader keeps the frame length intact. The
non-sparse RFID signals is provided. The unit interval sample
DFSA methods can achieve the highest system efficiency
is used to determine the source signal’s base vector. To pre-
when the length of the initial frame, or the slot number in the
vent the emergence of a mixed matrix, a comparable signal
initial frame, is the same as the number of tags. The frame
with the least interference is extracted afterwards. According
length for reading must be selected based on the number of
to the throughput study, the suggested approach outperforms
tags which are typically unknown to the reader and must be
the standard blind source separation (BSS) and ALOHA
estimated using an estimation method for optimal results. The
[35] algorithms. Six tags could be precisely separated using
two key contributions of the present work are:
three antennas. The approach would be useful in cases when
• A novel DFSA anti-collision algorithm for reading a collision detection is necessary and signal recovery is not
large number of tags by employing multiple receiver completely sparse.
antennas for applications in precision livestock farming. The state-of-the-art anti-collision algorithms [36] are pri-
• A MATLAB based simulation testbed for evaluating marily concerned with lowering collision probability, and
MIMO RFID anti-collision algorithms. as such, they suffer from the problem of large idle periods.
The paper is organized such that: Section II details the Su et al. [37] offer a collision-tolerant dynamic framed slotted
state-of-the-art techniques addressing the underlying prob- ALOHA approach (which they named CE-DFSA) that may
lem, while Section III explains the proposed technique. recognize several tags in every slot to shorten the identifica-
Section IV is about the simulation setup, the results achieved tion time of a tag. The orthogonal Walsh Sequence is used to
and a brief discussion along with highlighting the potential identify several tags without the need of a spread spectrum.
future work, Section V concludes the paper. Numerous tags with orthogonal WS respond once the RFID
reader sends a query. The system’s throughput, when bearing
II. LITERATURE REVIEW in mind the discrepancy in slot duration, is not a suitable
A prominent topic of study in MIMO RFID systems involve estimate of the anti-collision performance in terms of the
improvements in the RFID reader’s bulk reading speed identification time.
through advances in collision recovery algorithms when sig- Tag collisions in multi-tag situations can decrease through-
nals from many tags interfere [30]. To deal with tag collisions put performance of RFID standards like the EPCglobal 1
in multi-antenna RFID systems, many collision recovery [38]. Because the tags at any one moment are of an unknown
methods have been exploited [12]. Most of these methods number, they are always expected to be more than what the
employed passive tags according to the EPCglobal Class-1 frame length allows, which typically must be same as the
Generation-2 standard protocol, with simulations employing number of identifiable tags. This degrades the performance
a double Rayleigh fading channel. Backscatter link frequency of the anti-collision algorithms. By optimizing the estimation
VOLUME 11, 2023 130199
 M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

process for the quantity of tags an RFID reader’s throughput

can be significantly enhanced. Receiver algorithms designed
for a single antenna have been more commonly in use for
a long time, but they are hard to adapt to MIMO systems
due to the low SNR values caused by tag signal collisions,
which degrades the overall system performance. Deng et
al. [39] developed a multi-dimensional estimation approach
for MIMO based systems that utilizes spatial diversity at
the receiver’s several antennas to estimate tag quantity.
The colliding tag signals are kept as high-dimensional vec-
tors, which allows tag quantity estimates to be treated as
clusters of high-dimensional data. The modeling approach
allows for strengthening through separation by increments
in distance between clusters for acceptable SNR in multiple
backscattering channels. A density-based-spatial-clustering
of applications with noise (DBSCAN) algorithm was pro- FIGURE 3. Access efficiency of a typical DFSA system.

posed to estimate tags reliably, by being incorporated into

the high-dimensional space. With rapid advancements in soft-
ware defined radio (SDR) technologies, the development of where 1l (t) denotes the lth tag antenna’s load reflection
MIMO based RFID readers have gathered pace. With high coefficient at time t. Several load reflection coefficients cor-
density tag reading requirements such as those for remote respond to the different tag antennas in this case. A diagonal
sensing of crowds, MIMO systems can play a significant signaling matrix with coefficients of load reflections which
role in enhancing system throughput and improving reading are asymmetrical is often produced with if pre-determined
reliability. space time codes are employed in circuit design to represent
reflection coefficients of different tag antennas. We examine
III. METHODOLOGY a general MIMO RFID channel consisting of a tag with L
The matrix form of received signals in their baseband form at antennas, whereas M transmitting antennas are employed at
time t is represented [40] as: receiver for query signals to be sent to the tag, and N receiving
antennas for signals to be received from the tag. As explained
1 ∞ ∞ b
Z Z
r (t) = H (τb ; t) S (t) H f τf ; t x t − τb − τf earlier, the transmit signals act as a circuit charging method in



2 −∞ −∞ a passive RFID system because they send (identical) unmod-
× dτb dτf + w (t) (1) ulated waves to the tag. Additionally, we make a case for
where r(t) is vector of signals received at the reader receivers considering the statistical properties of a reader transmission
having dimensions N × 1, w(t) is a vector representing using a single antenna versus employing multiple antennas
the corresponding received noise components having dimen- are expected to be the same if the transmitting energy is
sions N × 1, and the vector x(t) having dimensions M × normalized to unity. This is because the transmitter antenna to
1 is representations of the unmodulated transmitted signal or the tag antenna channel gains are based on complex Gaussian
carrier signal/wave. The matrix H f τf ; t having dimensions

distribution. For the sake of simplicity, we will consider
L × M and H b (τb ; t) having a size of N × L, represent the 1 × N × L channel in the remainder of this paper and refer
forward and backscattering link impulse responses. The time to it simply as the channel N × L. We will also disregard
varying modulation and subsequent coding of the carrier sig- the time index t in the model going forward, also for the
nal by L th antenna RF tag is described by the signaling matrix sake of presentation simplicity. We let s = (s1 , . . . , sL )T =
S(t), having L × L dimensions. S(t) can take a number of (11 , . . . , 1L )T considering that L transmission symbols are
distinct shapes depending on the various tag circuit designs: transmitted simultaneously from the L antennas of RFID
The identical signaling matrix is one form represented as: tags. Hence, the received signal at any instant of time can be
expressed in vectorial form as [40]:
S (t) = 1 (t) I (2)
r = Hs + w (4)
where I stands for the L × L identity matrix and 1(t) refers
to the reflection coefficient of the RF tag load. In this similar there r = (r1 , . . . , rN )T is the received signal vector, noise
situation, the RF tag’s antennas all have the same reflection vector is represented by w where each of its identical elements
coefficient 1(t). The diagonal signaling matrix is another represents independently distributed Gaussian noise having
type shown below: zero mean and variance of unity, and H denotes the N × L
channel matrix with N rows and L columns.
11 (t)
 
The optimal frame size must be determined based on the
S (t) =  ..
. (3)
 
 number of tags and it must not be too large or there will be
1L (t) a lot of idle slots. Also, if the frame size is smaller than the

130200 VOLUME 11, 2023

M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

Algorithm 1 Proposed MIMO-DFSA Technique

1. Read numTotalTags
2. Read frameSizeIncr
3. Read initialFrameSize
4. Read numRcvAntenna
5. numBatches = numTotalTags/frameSizeIncr
6. numTags = initialFrameSize
7. Repeat numBatches times
a. Inititalize Time = 0
b. Increment numTags by frameSizeIncr
c. Repeat until numTags = 0
i. Create array uncollidedTagIDs
ii. Repeat for all antennas
1. Randomly generate slots
2. Initialize i, j = 1
3. Repeat numTags times
a. Initialize flag = 0
b. Repeat numTags times
i. k = i + j
ii. if slot(i) == slot(k)
1. Increment flag by 1
iii. increment j
c. if flag <> 0, increment collisions, else update uncollidedTagIDs with i
d. increment i
4. Collect uncollidedTagIDs
iii. collisions = numTags – uncollidedTagIDs
iv. Compute the total time to read numTags
v. Update numTags = collisions

number of tags to be read, the probability of collision will be

higher.
Figure 3 represents the reading efficiency for various tags
when a particular frame size is selected. For a frame size F,
the probability that a tag chooses a particular slot is 1/F and
the probability of choosing any other slot is 1 − 1/F. Thus,
for Q number of tags, the success slot probability will be:
1 Q−1
 
1
Psuccess = 1− (5)
F F
The probability of a tag that is successfully read in a given
frame of size F is:
1 Q−1
 
FPsuccess = 1 − (6) FIGURE 4. Structure of a single frame with collision in Slot 2.
F
The successful number of tags in a given frame will be:
efficiency will be at F = Q. This can also be observed in
1 Q−1
 
Fsuccess = Q 1 − (7) Figure 3.
F During the inventory round of EPC global class 1 gen-
Thus, the efficiency of access can be denoted as: eration 2 standard, the reader issues a QUERY command
to all the tags. Upon reception, every tag selects a random
1 Q−1
 
Q
ηsuccess = 1− (8) number between 0 to 2F −1. All tags who have selected num-
F F ber 0 responds to the QUERY command with a 16-bit ID
where ηsuccess is a single variable function of the number number called the RN16. The tags that have selected random
of tags F, if the frame size Q is fixed. From equation 8, numbers other than 0 shall run a counter and respond on
we can deduce that for optimum performance, the maximum their respective turns. If RN16 is successfully identified for a

VOLUME 11, 2023 130201

 M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

particular tag, the reader responds with an acknowledgement

(ACK) to it and the tag will respond back with its electronic
product code (EPC) and a cyclic redundancy check (CRC).
The reader then sends a QUERY REP command to all the
remaining tags who will then decrement their counters by
1 and the one which has a 0 will respond with its RN16 ID.
Collision can occur when multiple tags select the same RN16
and respond in the same slot such as Slot 2 in Figure 4.
In DFSA, a frame’s initial number of slots is set, and tags
are assigned at random. The reader then reads in the assigned
time slots (Figure 2) and if no collision is detected, the reading
process terminates, but if collisions are detected, the process
is repeated by using the number of idle slots to calculate the
new frame size, occurring collisions and successful reads.
Therefore, for every upcoming read cycle, the reader selects
whether to increase, reduce, or keep the previous frame size.
For instance, if the quantity of colliding slots exceeds the
FIGURE 5. MIMO DFSA performance for receiver with 1, 2 and 4 antennas
higher threshold, frame size is increased; and if the number and 500 tags.
of colliding slots is less than the lower threshold, the frame
size is decreased by the reader. Otherwise, the quantity of
colliding slots is within limits, and the frame size is kept
intact. The DFSA methods can achieve the highest system
efficiency when the number of tags are same as the length of
the initial frame, or the slot number in the initial frame. The
starting frame length should be chosen in accordance with
the number of tags, which is typically unknown to the reader,
to achieve highest efficiency.
Algorithm 1 details the proposed DFSA technique for
anticollision in MIMO RFID readers with N receive antenna.
Assuming that the total number of RFID tags are known, the
algorithm reads the tags in batches where the number of tags
in the next batch are incremented by 10. For example, let the
total quantity of tags be 500 and we start with 10 tags. The
first batch will have 10 tags, the second 20, the third 30, and
so on. We also assume that the number of time slots are same
as the quantity of tags. The algorithm begins by randomly
assigning time slots to each tag which is equivalent to the FIGURE 6. MIMO DFSA performance for receiver with two antennas,
500 tags and various interference probabilities.
reader reading the assigned time slots. The algorithm then
reads the time slot of each tag and identifies those tags that
have identical numbers in the time slots. This is equivalent the performance of the proposed MIMO setup with that of
to detecting collisions in the real scenario. The algorithm traditional setup that uses a single antenna. The performance
records the number of collisions and the tag numbers of was measured in terms of the time taken to read a certain num-
successful transmissions. The algorithm then assigns random ber of RFID tags. The proposed simulator takes the number
slots to the unsuccessful tags for the next antenna and this of tags as the input, along with the number of antennas to be
process is repeated for each antenna. If there still are tags that used and the interference probability value. The maximum
had encountered collisions, the algorithm starts again for each throughput of 1/e [41] is achieved in slotted aloha protocol
antenna until all the tags are read. when frame size and number of tags are equal. The simulator
allows for adjusting interference probability Pi to be selected
IV. SIMULATION SETUP AND RESULTS between 0 and 1. It is the probability of missing a set of
The simulations for particular cases of application of the tags due to physical interference. Time to read N tags with
proposed DFSA technique were performed using a Lenovo interference in frame size F will be TN = (e ∗ N )/(1 − Pi ).
Legion PC with an Intel Core i7-11700 2.5 GHz, a 16GB Frame size can also be restricted if evaluation of access
RAM and a 2 TB disk space running a Windows operating efficiency is desired.
system. The proposed MIMO RFID testbed is designed in Figure 5 compares the tag reading performance of DFSA
MATLAB. We ran several simulations using the proposed algorithm for 1, 2 and 4 antennas, and with a probability of
algorithm. The objective of these simulations was to compare interference as 0.4. It can be observed that the time to read

130202 VOLUME 11, 2023

M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

FIGURE 7. Performance gains using multiple receiver antennas.

FIGURE 9. Time to read 1000 tags using 2 antennas and fixed frame
lengths.

tags grow beyond 256 and 512, the performance deteriorates

significantly for a smaller frame size and performance well in
case of a larger frame size. The optimum performance will be
when the frame size is changed per frame. For comparison,
Figure 9 shows the results of robust MIMO DFSA algorithm,
where the reading time has been significantly reduced despite
the fixed frame sizes. The read time for up to 500 tags is
almost the same for all frame sizes. However, since the frame
size is fixed, as the number of tags increases, the interference
will increase as well as the idle slots per frame resulting in
FIGURE 8. Time to read 1000 tags using a single antenna and fixed frame
suboptimal performance but still better than a single antenna
lengths. system.

the total number of tags with DFSA over a single receive V. CONCLUSION
antenna system is around 2.25 seconds while that with a When there are multiple tags in an RFID reader’s range,
two-antenna system is a little over 1 second, and that with it is difficult for the reader to differentiate between sev-
a four-antenna system is a little below 1 second. This shows eral signals sent by the tags. This is a significant problem
that the use a two-antenna system or a four-antenna one is in small ruminant farms where herd density is very high,
preferable to the employment of a single-antenna system. and a high reading rate is desired to avoid reading errors.
Moreover, we observe also that the performance improve- A MIMO technology-based ALOHA based anti-collision
ment for a two-antenna system over a single antenna one algorithm was proposed for throughput enhancement in high
is more significant than the improvement of a four-antenna tag population scenarios particularly large farms. With the
system over a two-antenna one. This is to be expected because rapid advancements in precision livestock farming technolo-
as the number of antennas increase beyond two, the number gies and elevated requirements, multi-antenna RFID systems
of rounds for reading the same number of tags cannot go can significantly improve the process of reading large number
lower than one or two and hence, the improvement is not that of tags if the anti-collision algorithms are adapted to such
significant. configurations. The proposed algorithm outperforms single
The performance of an RFID system in the presence antenna systems in the presence of interference and enhances
of various interference values, with two receive antennas, reliability of an RFID reader during bulk tag reading process
is demonstrated in Figure 6. The simulation results for up along with significant throughput gains. The future work
to 500 tags show that the reading time using two antennas includes development of an Artificial Neural Network (ANN)
and dynamic frame sizes significantly improves compared based estimation technique. The system will the received sig-
to a single receive antenna. Moreover, as the probability of nal strength of the backscatter query response for estimating
interference increases, the time to read tags increases which the number of tags to optimally select the frame size.
is demonstrated by the upward trend of the curves. Figure 7
shows the performance gains. ACKNOWLEDGMENT
Figure 8 shows the reading times if the frame sizes are The authors would like to thank the Smart Laboratory,
fixed. The results for a single antenna system show that Department of Computer Engineering College of Computer
for a frame size fixed at 256 and 512, as the number of Information Systems, Deanship of Scientific Research and

VOLUME 11, 2023 130203

 M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

Science and Technology Unit at Umm Al Qura University for [24] C. He, X. Chen, Z. J. Wang, and W. Su, ‘‘On the performance of MIMO
their logistic support. RFID backscattering channels,’’ EURASIP J. Wireless Commun. Netw.,
vol. 2012, no. 1, p. 357, Dec. 2012.
[25] L. Arjona, H. Simon, and A. Ruiz, ‘‘Energy-aware RFID anti-collision
REFERENCES protocol,’’ MDPI Sensors, vol. 18, no. 6, p. 1904, Jun. 2018.
[1] M. C. Vuran, A. Salam, R. Wong, and S. Irmak, ‘‘Internet of Underground [26] M. Shakiba, A. Zavvari, and E. Sundararajan, ‘‘Fitted dynamic framed
Things in precision agriculture: Architecture and technology aspects,’’ Ad slotted Aloha anti-collision algorithm in RFID systems,’’ in Proc. 5th Int.
Hoc Netw., vol. 81, pp. 160–173, Dec. 2018. Conf. Inf. Technol. Multimedia (ICIMU), Nov. 2011, pp. 1–6.
[2] U. Shafi, R. Mumtaz, J. García-Nieto, S. A. Hassan, S. A. R. Zaidi, and [27] Z. Jiang, B. Li, M. Yang, and Z. Yan, ‘‘LC-DFSA: Low complexity
N. Iqbal, ‘‘Precision agriculture techniques and practices: From consider- dynamic frame slotted Aloha anti-collision algorithm for RFID system,’’
ations to applications,’’ Sensors, vol. 19, no. 17, p. 3796, Sep. 2019. Sensors, vol. 20, no. 1, p. 228, Dec. 2019.
[3] P. Visconti, R. de Fazio, R. Velázquez, C. Del-Valle-Soto, and [28] K. C. Shin, S. B. Park, and G. S. Jo, ‘‘Enhanced TDMA based anti-collision
N. I. Giannoccaro, ‘‘Development of sensors-based agri-food traceability algorithm with a dynamic frame size adjustment strategy for mobile RFID
system remotely managed by a software platform for optimized farm readers,’’ Sensors, vol. 9, no. 2, pp. 845–858, Feb. 2009.
management,’’ Sensors, vol. 20, no. 13, p. 3632, Jun. 2020. [29] R. P. B. Mota and D. M. Batista, ‘‘A dynamic frame slotted Aloha anti-
[4] L. Ruiz-Garcia and L. Lunadei, ‘‘The role of RFID in agriculture: Appli- collision algorithm for the Internet of Things,’’ presented at the 29th Annu.
cations, limitations and challenges,’’ Comput. Electron. Agricult., vol. 79, ACM Symp. Appl. Comput., Gyeongju, Republic of Korea, Mar. 2014, doi:
no. 1, pp. 42–50, Oct. 2011. 10.1145/2554850.2555003.
[5] G. Caja, A. Castro-Costa, A. A. K. Salama, J. Oliver, M. Baratta, C. Ferrer, [30] Y. Zeng and H. Ding, ‘‘A physical-layer UHF RFID tag collision resolution
and C. H. Knight, ‘‘Sensing solutions for improving the performance, based on Miller code,’’ Wireless Commun. Mobile Comput., vol. 2021,
health and wellbeing of small ruminants,’’ J. Dairy Res., vol. 87, no. S1, Apr. 2021, Art. no. 6636846.
pp. 34–46, Aug. 2020. [31] H. Salah, H. A. Ahmed, J. Robert, and A. Heuberger, ‘‘Multi-antenna
[6] R. Weinstein, ‘‘RFID: A technical overview and its application to the UHF RFID reader utilizing stimulated rate tolerance,’’ IEEE J. Radio Freq.
enterprise,’’ IT Prof., vol. 7, no. 3, pp. 27–33, May 2005. Identificat., vol. 1, no. 2, pp. 124–134, Jun. 2017.
[32] A. Abbasian and M. Safkhani, ‘‘CNCAA: A new anti-collision algorithm
[7] J. Landt, ‘‘The history of RFID,’’ IEEE Potentials, vol. 24, no. 4, pp. 8–11,
using both collided and non-collided parts of information,’’ Comput. Netw.,
Oct. 2005.
vol. 172, May 2020, Art. no. 107159.
[8] R. Want, ‘‘An introduction to RFID technology,’’ IEEE Pervasive Comput.,
vol. 5, no. 1, pp. 25–33, Jan./Mar. 2006, doi: 10.1109/MPRV.2006.2. [33] D. K. Klair, K.-W. Chin, and R. Raad, ‘‘A survey and tutorial of RFID
anti-collision protocols,’’ IEEE Commun. Surveys Tuts., vol. 12, no. 3,
[9] K. Domdouzis, B. Kumar, and C. Anumba, ‘‘Radio-frequency identifi-
pp. 400–421, 3rd Quart., 2010.
cation (RFID) applications: A brief introduction,’’ Adv. Eng. Informat.,
[34] X. Cheng and C. Liu, ‘‘Research on RFID collision detection algorithm
vol. 21, no. 4, pp. 350–355, Oct. 2007.
based on the underdetermined blind separation,’’ in Proc. 4th Int. Conf.
[10] B. Nath, F. Reynolds, and R. Want, ‘‘RFID technology and applications,’’
Machinery, Mater. Comput. Technol. (ICMMCT), 2016, pp. 1292–1299.
IEEE Pervasive Comput., vol. 5, no. 1, pp. 22–24, Jan./Mar. 2006.
[35] S. Feng, M. Wang, J. Yan, Y. Zhu, and Z. Li, ‘‘Independent component
[11] F. Klaus, RFID Handbook: Fundamentals and Applications in Contactless
analysis based tag anti-collision algorithm in multi-antenna radio fre-
Smart Cards and Identification. Hoboken, NJ, USA: Wiley, 2003.
quency identification,’’ in Proc. 5th Int. Conf. Inf. Sci. Technol. (ICIST),
[12] M. Alotaibi, M. Murad, S. A. H. Alhuthali, F. R. Al-Osaimi, and
Apr. 2015, pp. 519–524.
F. Aldosari, ‘‘MIMO radio frequency identification: A brief survey,’’ Sen-
[36] S. Ullah, W. Alsalih, A. Alsehaim, and N. Alsadhan, ‘‘A review of tags
sors, vol. 22, no. 11, p. 4115, May 2022.
anti-collision and localization protocols in RFID networks,’’ J. Med. Syst.,
[13] G. Deak, K. Curran, and J. Condell, ‘‘A survey of active and passive indoor
vol. 36, no. 6, pp. 4037–4050, Dec. 2012.
localisation systems,’’ Comput. Commun., vol. 35, no. 16, pp. 1939–1954,
[37] J. Su, Z. Sheng, and L. Xie, ‘‘A collision-tolerant-based anti-collision
Sep. 2012.
algorithm for large scale RFID system,’’ IEEE Commun. Lett., vol. 21,
[14] V. Chawla and D. Ha, ‘‘An overview of passive RFID,’’ IEEE Commun.
no. 7, pp. 1517–1520, Jul. 2017.
Mag., vol. 45, no. 9, pp. 11–17, Sep. 2007.
[38] P. Šolic, J. Maras, J. Radic, and Z. Blaževic, ‘‘Comparing theoretical and
[15] P. V. Nikitin and K. V. S. Rao, ‘‘Performance limitations of passive UHF experimental results in Gen2 RFID throughput,’’ IEEE Trans. Autom. Sci.
RFID systems,’’ in Proc. IEEE Antennas Propag. Soc. Int. Symp., Jul. 2006, Eng., vol. 14, no. 1, pp. 349–357, Jan. 2017.
pp. 1011–1014.
[39] W. Deng, Y. Xia, Z. Li, R. Pu, and W. Pei, ‘‘A high-dimensional collided
[16] G. De Vita and G. Iannaccon, ‘‘Design criteria for the RF section of UHF tag quantity estimation method for multi-antenna RFID systems,’’ IEEE
and microwave passive RFID transponders,’’ IEEE Trans. Microw. Theory Commun. Lett., vol. 25, no. 1, pp. 132–136, Jan. 2021.
Techn., vol. 53, no. 9, pp. 2978–2990, Sep. 2005.
[40] C. He, X. Chen, Z. J. Wang, and W. Su, ‘‘On the performance of MIMO
[17] D. Yeager, F. Zhang, A. Zarrasvand, and B. P. Otis, ‘‘A 9.2 µA gen 2 com- RFID backscattering channels,’’ EURASIP J. Wireless Commun. Netw.,
patible UHF RFID sensing tag with −12 dBm sensitivity and 1.25 µVrms vol. 2012, no. 1, pp. 1–15, Dec. 2012.
input-referred noise floor,’’ in IEEE Int. Solid-State Circuits Conf. (ISSCC) [41] J. Yu, P. Zhang, L. Chen, J. Liu, R. Zhang, K. Wang, and J. An, ‘‘Sta-
Dig. Tech. Papers, Feb. 2010, pp. 52–53. bilizing frame slotted aloha-based IoT systems: A geometric ergodicity
[18] P. V. Nikitin and K. V. S. Rao, ‘‘Gain measurement of antennas using perspective,’’ IEEE J. Sel. Areas Commun., vol. 39, no. 3, pp. 714–725,
RFID,’’ in Proc. IEEE Int. Symp. Antennas Propag. (APSURSI), Jul. 2011, Mar. 2021.
pp. 1012–1015.
[19] A. K. Jones, S. Dontharaju, S. Tung, P. J. Hawrylak, L. Mats, R. Hoare,
J. T. Cain, and M. H. Mickle, ‘‘Passive active radio frequency identifi-
cation tags,’’ Int. J. Radio Freq. Identificat. Technol. Appl., vol. 1, no. 1,
pp. 52–73, 2006.
[20] J. D. Griffin and G. D. Durgin, ‘‘Multipath fading measurements for multi- MAJID ALOTAIBI received the Ph.D. degree from
antenna backscatter RFID at 5.8 GHz,’’ in Proc. IEEE Int. Conf. RFID,
The University of Queensland, Brisbane, QLD,
Apr. 2009, pp. 322–329.
Australia, in 2011.
[21] D. Kim, M. A. Ingram, and W. W. Smith, ‘‘Measurements of small-scale
fading and path loss for long range RF tags,’’ IEEE Trans. Antennas
He is currently a Professor with the Depart-
Propag., vol. 51, no. 8, pp. 1740–1749, Aug. 2003. ment of Computer Engineering, Umm Al-Qura
[22] J. D. Griffin and G. D. Durgin, ‘‘Gains for RF tags using mul- University, Makkah, and the Founding Member
tiple antennas,’’ IEEE Trans. Antennas Propag., vol. 56, no. 2, of the Smart Laboratory. His current research
pp. 563–570, Feb. 2008. interests include mobile computing, mobile and
[23] A. Bekkali, S. Zou, A. Kadri, M. Crisp, and R. V. Penty, ‘‘Performance sensor networks, wireless technologies, the IoT in
analysis of passive UHF RFID systems under cascaded fading channels healthcare, smart cities, ad hoc networks, com-
and interference effects,’’ IEEE Trans. Wireless Commun., vol. 14, no. 3, puter networks (wired/wireless), RFID, antennas and propagation, radar, and
pp. 1421–1433, Mar. 2015. nano electronics.

130204 VOLUME 11, 2023

M. Alotaibi et al.: Anti-Collision Algorithm for Identification in Precision Agriculture Applications

MOHSIN MURAD received the M.S. degree in FAISAL R. AL-OSAIMI received the B.Sc. degree
computer systems engineering from the Univer- in electrical and computer engineering from Umm
sity of Engineering and Technology, Peshawar, Al-Qura University, Saudi Arabia, in 2000, the
Pakistan, in 2011, and the Ph.D. degree in telecom- M.E. degree in computer systems engineering
munications engineering from the University of from The University of Queensland, Australia,
Malaga, Malaga, Spain, in 2022. in 2005, and the Ph.D. degree in the area of 3D
He is currently a Lecturer with the Department face biometrics from The University of Western
of Computer Engineering, Umm Al-Qura Univer- Australia, in 2010.
sity, Makkah, and a Senior Researcher with the He is currently a Faculty Member of the Depart-
Smart Laboratory. His research interests include ment of Computer Engineering, Umm Al-Qura
wireless communication, underwater communication, the Internet of Things, University. His research interests include computer vision, biometrics, and
embedded systems, and wireless sensor networks. pattern recognition.

FAHD ALDOSARI received the B.Sc. degree in

computer engineering from the King Fahd Univer-
sity of Petroleum and Minerals, Dhahran, Saudi
Arabia, the M.Sc. degree in mobile and satellite
communications and the Ph.D. degree in computer
networks from Bradford University.
He had served as the Dean of the Computer
IMRAN ALI TASADDUQ received the B.E. degree in electrical engineering
and Information Systems College and the Dean
from NED University, Pakistan, the M.S. degree in systems engineering
of IT with Umm Al-Qura University, Makkah,
from King Fahd University, Saudi Arabia, and the Ph.D. degree from the
Saudi Arabia. He is currently a Faculty Member
Department of Electrical and Computer Engineering, Western University,
with the Department of Computer Engineering, Umm Al-Qura University.
London, ON, Canada.
His research interests include cloud computing, computer networks, and
He was with the University of Ottawa; Research Institute, King Fahd
digital transformation.
University; Sir Syed University of Engineering and Technology; and FAST—
National University, Pakistan. He is currently a Professor and an ABET
Coordinator of the Department of Computer Engineering, Umm Al-Qura
University, Makkah, Saudi Arabia. Besides authoring and coauthoring two ALI MUHAMMAD RUSHDI (Life Senior Mem-
books, one book chapter, he has produced close to 40 research publications ber, IEEE) was born in Port Said, Egypt, in May
in international journals and conferences and has served on the technical 1951. He received the B.Sc. degree in electrical
committees for various international conferences. His research interests engineering from Cairo University, Giza, Egypt,
include the various domains of wireless communications, such as OFDM in 1974, and the M.S. and Ph.D. degrees in elec-
and GFDM, underwater acoustic communication, optimization techniques, trical engineering from the University of Illinois
facility layout design, and accreditation issues in engineering education. at Urbana-Champaign (UIUC), USA, in 1977 and
1980, respectively. In 1974, he was appointed
as a Demonstrator and an Instructor with the
Department of Electronics and Electrical Com-
munications, Cairo University. From 1976 to 1980, he was a Research
Assistant and a Consultant with the Electromagnetics Laboratory, UIUC.
Since 1980, he has been with King Abdulaziz University (KAU), Jeddah,
Saudi Arabia, where he is currently a Distinguished Adjunct Professor in
electrical and computer engineering and the former Head of the Computer
Engineering Programs. At KAU, he structured and taught a great number
SHAKIR A. H. ALHUTHALI received the master’s of graduate and undergraduate courses, supervised over 85 M.S. and Ph.D.
degree in electric and computer engineering from theses, and contributed significantly to accreditation activities. He has served
King Abdulaziz University, Jeddah, Saudi Arabia, (and is serving) as an Editor or an Associate Editor for several prestigious
in 2014, where he is currently pursuing the Ph.D. scientific journals. His research interests and contributions over the past
degree. 45 years spanned the areas of electromagnetic communications, computer
He has been a Researcher with Umm Al Qura engineering, reliability, digital design, engineering pedagogy and education,
University, Makkah, Saudi Arabia, since 2008. switching networks, Boolean algebras and equations, engineering design,
He has served as a Laboratory Instructor for dimensional analysis, diagnostic testing, ecological modeling, futurology,
Computer Engineering Department and provided translation, inferential thinking, and innovative problem solving. He is an
tutorial classes for students. His research interests initiated member of the Honorary Societies: Eta Kappa Nu and Phi kappa
include computer vision, the Internet of Things, and reliability by using Phi.
Boolean equations.

VOLUME 11, 2023 130205