Received 15 November 2024, accepted 3 December 2024, date of publication 9 December 2024,

date of current version 17 December 2024.

Digital Object Identifier 10.1109/ACCESS.2024.3512785

Fuzzy Logic Driven Load Balanced Channel

Allocation Algorithm for Enhancing Channel
Utilization and Throughput in Large
Heterogeneous IEEE 802.11ah Network
R. NANDHINI AND R. RADHA
School of Computer Science and Engineering, Vellore Institute of Technology, Chennai 600127, India
Corresponding author: R. Radha ([email protected])

ABSTRACT A wireless sensor network typically comprises of heterogeneous nodes with varying sampling
rates and sample sizes. IEEE 802.11ah supports up to 8192 sensor nodes, offering a data rate of
approximately 16 Mbps from an individual access point. It utilizes a feature called Restricted Access Window
(RAW) in the MAC protocol; a limited number of nodes are assigned to access the channel at a designated
time to reduce contention. However, most of the existing mechanisms do not account for this heterogeneity,
and they randomly group the nodes. In this work, we introduce a Load Balanced Channel Allocation (LBCA)
algorithm for a large heterogeneous sensor network. This algorithm collects the sampling interval and sample
size from the node and identifies the appropriate group based on the fuzzy logic approach. We calculate the
grouping probability (GP) for each RAW and assign the node to the RAW with the highest GP. The number
of nodes, the standard deviation of RAW utilization ratio among the groups and the average transmission
efficiency rate were considered as the input parameters for the fuzzy logic. We simulated this work using
ns-3 simulator and evaluated the performance against random grouping in terms of improvement in channel
utilization of around 7.65%, throughput by 1.37%, decrease in collision rate by 16.42% and minimization
of delay by about 8.57%. These results highlight the effectiveness of the LBCA algorithm in achieving load
balance along with efficient channel utilization.

INDEX TERMS Channel utilization, transmission efficiency, grouping, heterogeneous, IEEE 802.11 ah,
traffic.

I. INTRODUCTION single access point [1]. In sensor networks, nodes can be

IEEE 802.11ah also known as Wi-Fi HaLow, is a wireless in either active or sleep mode. When a node is in sleep
networking standard crafted specifically to address the mode, it does not listen to the network and conserves
specific demands of the Internet of Things (IoT). The power. However, this can lead to latency and delays in
standard operates in unlicensed sub-1 GHz frequency bands communication when the node wakes up and needs to
and is engineered to provide low-power, cost-effective, and transmit data. To tackle this issue, IEEE 802.11 ah (HaLow)
long-range wireless communication. The significance of introduces the concept of a Restricted Access Window
IEEE 802.11ah is to provide a standardized communication (RAW). The RAW is a predetermined period during which
protocol that allows IoT devices to connect to the internet devices can wake up and transmit data as discussed in [2].
seamlessly and communicate with each other. It supports The protocol allows only a subset of devices in the
around 8000 sensor nodes that can be associated under a network to transmit data during RAW. This helps to reduce
the probability of collisions and increase the efficiency of
The associate editor coordinating the review of this manuscript and accessing the channel. Devices that are not in the RAW must
approving it for publication was Huan Zhou . remain in sleep mode conserving power. Channel time is
2024 The Authors. This work is licensed under a Creative Commons Attribution 4.0 License.
185838 For more information, see https://creativecommons.org/licenses/by/4.0/ VOLUME 12, 2024
R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

divided into multiple beacon intervals where each beacon

interval consists of a contention-based time window and a
structured RAW period as shown in Figure 1. Structured
RAW period was further divided into multiple fixed RAW
groups where limited number of assigned nodes contend
in each slot [3], where contention is managed by limiting
the number of stations assigned to specific slots, although
contention still exists as stations access the channel within
those slots using the EDCA protocol. Any node in the
network can access the contention-based window using the
standard 802.11 Enhanced Distributed Channel Access and
Distributed Coordination Function (EDCA/DCF) protocol.
Each sensor node is assigned an unique AID (Association
identifier) during the association phase and the ID is relative
to the assigned group. The nodes wake up and listen to the
Raw Parameter Set element (RPS) which is carried by beacon
to know its RAW timing and its duration. The basic frame FIGURE 1. Beacon period.
format for the RPS is given below in Figure 2. The RPS
advertises the start and end association ID (AID), as well as
the start time, and duration of the RAW. Furthermore, each
RAW consists of one or more equal duration slots, among
which the RAW group’s stations are uniformly distributed
(using round-robin assignment). The number of slots, slot FIGURE 2. RPS frame.
format, and slot length are all included in the RPS information
element, and together they constitute the RAW slot duration.
Information retrieved from the RPS frame can be used by Section IV demonstrate the outputs obtained through the
sensor nodes to know their wake-up time. simulation of algorithm in a network simulator. Finally,
The mapping of the station to the RAW slot is done using Section V concludes the paper by describing the procedure
the Eq.(1) mentioned below: carried out to develop the enhanced LBCA algorithm which
performs better when compared with random method.
SNislot = (SN + Hoffset ) mod N (1)
ith
where SNislot is the index of RAW slot among which the II. LITERATURE SURVEY
station is being mapped. N is the count of slots in each RAW. The survey in this section provides an overview and insight
Hoffset is the offset value for improving fairness and it is into various methods developed in Wi-Fi Halow and in its
usually the two bytes from the least significant position of the major MAC protocol-RAW. The subsections give a brief note
checksum field from the beacon frame and SN is the index or on the techniques incorporated in RAW optimization under
Association ID (AID) of the station [4]. static and dynamic conditions, grouping process utilized
The 802.11ah specification makes no mention about the to enhance network performance, approaches involved to
way of grouping and setting up of the RAW parameters. mitigate hidden terminal issues, how network performance
Hence grouping of sensors were done randomly without indices are improved while considering heterogeneous traffic
paying any heed to their transmission rates. Furthermore, past and illustrating the limitations in all the methods.
research had revealed that the best RAW configuration is
influenced by several network-related characteristics. Hence, A. STATIC vs. DYNAMIC RAW OPTIMIZATION
in this paper, an algorithm is developed with the assumption The study by [5] specifically considering low-intensity
of fixed number of RAW each with the single long RAW IoT applications was focussed on optimizing the PRAW
slot as described in [5]. This approach improves the station parameters, utilization of single RAW slot and observation
grouping by considering the node’s required transmission of crossing effect help in reducing CTC. By performing
time, transmission efficiency rate, and RAW utilization rate modification in RAW period duration or in optimal number of
of the RAW groups. This algorithm is designed to meet the groups based on traffic pattern helps to minimize CTC, delay
heterogeneous traffic requirements of the IEEE 802.11ah use and energy requirements, but this approach is limited to sta-
cases [6] & [7]. ble, low-intensity networks. The limitation in standard IEEE
The structure of the research paper is as follows: Section II 802.11ah is resolved in [8] by developing the algorithm for
discusses the prior works done in the existing literature with dynamic assignment of RAW based on the network traffic and
the motive to enhance network performance. Section III active stations. The results demonstrate improvement in aver-
provides a detailed explanation of the proposed method age delay(0.43sec), throughput (23kbps), maximum PDR
and overview of the channel allocation algorithms utilized. (98%) and jitter (80sec) compared to static RAW. Though

VOLUME 12, 2024 185839

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

it lacks adaptive load-balancing and traffic differentiation average of 6% and 4% over other grouping methods but may
capabilities. The machine learning-based model is developed lack flexibility for varying traffic loads.
in [9] to optimize RAW and station grouping parameters The sector-based device grouping scheme utilizing a
for the network operating with dynamic traffic conditions. spatial orthogonal access mechanism is introduced in [17].
The method has significant improvement in throughput, The result showcase substantial improvement in through-
minimization in delay and execution time of about 13.53%, put, reduction in the system delay and frame collisions
18.01%, 30.32% over TAROA and 52.88%, 70.88%, 41.65% when compared with conventional grouping mechanisms
over 802.11ah, but is contingent on predefined conditions. yet fixed sector allocations limit adaptability in dynamic
Additionally, authors in [10] developed an algorithm in which scenarios.
the scheduling for channel access were done with the help To solve the problem due to system irregularity a grouping
of RBT values of the station and the RAW is composed algorithm is developed in [18] utilizing the data rate and
of two phases. TRC-RBT-GC enhances system performance incorporating constraints like overhead with the motive to
by utilizing efficient claiming RAW/slot mechanisms and improve throughput and to optimize the performance but it
incorporation of load balancing among RAW slots. Yet may fall short amid rapid rate fluctuations in high-density IoT
limited by static scheduling that doesn’t account for rapid settings, resulting in performance limitations.
traffic changes.
The TAROA algorithm stated in [11] is designed to
adapt the RAW parameters based on traffic conditions, C. HIDDEN NODE MANAGEMENT
execution takes place at each target beacon transmission The challenges faced due to network non-uniformity in
time and parallelly communicate the RAW parameters to the IEEE 802.11ah IoT networks was addressed by sector-based
corresponding stations. It excels in terms of throughput for grouping in [19]. The strategy leverages adjustable cen-
dense networks, but may struggle with highly variable traffic. tral angles and incorporates three distinct criteria’s to
The spectral-based clustering introduced in [12] illustrates provides fair communication among STAs by mitigating
the grouping in RAW without relying on the geographic the hidden terminal problem, enhances overall network
data. The mechanism utilizes density levels, traffic patterns, throughput. The critical issues in Wi-Fi Halow standard
and propagation models, enhance the throughput, delay, and while supporting the sensor network applications is alleviated
collision rate over IEEE 802.11ah, K-Means Clustering and by designing hidden matrix-based regrouping algorithm
hidden matrix-based regrouping. However, its applicability in [20]. The detection of hidden node and regrouping
to dense IoT networks with diverse traffic characteristics of the same was done with the help of hidden node
remains limited. matrix. It shows superior performance than standard IEEE
802.11ah grouping in terms of minimizing hidden node
pairs and enhancing PS-Poll transmission times but not
B. GROUPING STRATEGIES catering to adaptive load-balancing for diverse traffic
A priority class-based grouping (PCG) algorithm was devel- profiles.
oped by [13] to solve the issues like starvation of lower- Another method is demonstrated in [21] combine two
priority stations. It makes use of the Service Characteristic scenarios: minimizing collisions and employing a grouping
field, making low priority STAs get adequate access, algorithm with prior knowledge of hidden devices to optimize
enhances overall throughput. PCG had better aggregate network performance but limiting scalability. The throughput
saturation output over random grouping mechanism, but has notable improvement of about 139% and 146% for 6 and
does not address dynamic traffic in dense, heterogeneous 8 groups. According to [22] the Max-Min fairness constraint
networks. Authors in [14] utilize the SOM neural network to was utilized to group stations being NP-hard in nature ACO
group devices based on their transmission rates using DGML is used to arrive at the best solution. It shows increase of
algorithm at a learning rate of 0.5, leveraging decapsulated 37% in minimum throughput per station, 40% in overall
PLCP after association phase. When compared with the throughput, and 11% reduction in count of hidden terminals
uniform grouping scheme, it improves average throughput over K-Means but the computational demands make real-time
and overcome a performance anomaly, though its complex application challenging.
training process limits real-time adaptability. To address the above said issue, [23] introduces Signal
The optimal traffic grouping algorithm (OTGA) along with Strength Assistant (SSA) grouping. In this method, multiple
the incorporation of branch and bound method is introduced Reference Nodes (RNs), each located at known positions,
in [15] to achieve optimal solution. Additionally heuristic were used to broadcast beacon signals. The simulation results
traffic sensor mapping algorithm is included to ensure a show that the SSA grouping cause reduction in hidden nodes
balance in traffic demands but struggles with real-time traffic thus reducing collision probability of about 0.05% over RG
variations. The expected traffic demands of a sensor is but lacks adaptability for shifting traffic. The approach pre-
taken by [16] to group them and the contention success sented in [24] focuses on handling the hidden node problem,
was analysed. Ensuring effective mitigation of biased traffic by incorporating regrouping strategy. MHPA initially identify
distribution, provides an improvement in channel utilization’s the nodes hampered by the hidden node problem, then

185840 VOLUME 12, 2024

R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

TABLE 1. Categorising Survey - Grouping Strategy.

perform regrouping. It proved to be very useful for dense net- D. HETEROGENEOUS TRAFFIC CONDITIONS
works having up to 8000 nodes achieving 97.8% reduction in Surrogate modelling named Model-based RAW Optimization
hidden pairs yet does not fully balance throughput and energy Algorithm (MoROA) was developed in [26] for optimizing
efficiency across varied traffic. The hidden terminal aware RAW configurations in case of dense heterogeneous environ-
grouping algorithm devised in [25] uses the SNIR parameter ment by utilizing the multi-objective optimization approach.
to mitigate hidden terminal. An analytical model is developed Specifically, it enhances throughput by 65% and achieves
to evaluate and analyse the effectiveness of this approach 96% lower energy consumption. ETAROA improves the
with existing methods in terms of throughput, latency, and throughput by 5.3% and energy efficiency by 32.4%.
energy efficiency. But the notable drawback is SNIR depen- The grouping model reported at [27] utilizes RNN along
dency may hinder adaptability for heterogeneous device with LSTM to perform the grouping in an intelligent way by
capabilities. exploiting the concept of cluster head formation. It provides

VOLUME 12, 2024 185841

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

enhancement in energy efficiency by up to 88.7%, increase grouping methodology by leveraging fuzzy logic to achieve
in network stability of up to 90.8%, improves network load balancing by optimizing the channel utilization while
scalability by up to 87.1%, reduction in congestion level of considering the heterogeneous traffic condition of the devices
up to 18.3%, and boosts Quality of Service in the range of in the network.
93.4%.
The algorithm for authentication control has been devel- III. PROPOSED SYSTEM
oped in [28] to minimize the association delay by considering Applications running in wireless sensor networks uses sen-
the best group size parameter as major constraint. Results sors with varying sampling rates and sample sizes depending
show notable reduction in total association time by factor of on several factors, including the type of sensor being used,
2.45 and 1.3 when compared with the group sizes of 30 and the application requirements, and the available network
10 respectively. Due to randomness, access windows of the resources. The sampling interval (SI) refers to the interval
nodes will lead to many hidden node collisions, which will between two consecutive samples measured by sensor node.
hamper the energy efficiency. It is typically represented in seconds (s) or minutes (min).
The load-balanced sensor grouping mechanism is pre- The sample size in a wireless sensor network (WSN)
sented in [29] aim to group sensors based on their het- depends on the type of data being collected and requirements
erogeneous traffic demands and were assessed using two of the application. The sample size can range from a few
parameters: Sampling rates and packet size. Notably, this bits to several kilobytes or even megabytes, depending on the
method achieved a 15% increase in channel utilization sensor and the data it is collecting. For example, a temperature
compared to random grouping but may falter in high-density sensor may collect data at a sample size of a few bytes per
traffic. The genetic algorithm-based grouping strategies were reading, while an image sensor may capture data at a sample
utilized in [30] for enhancing throughput and fairness in RAW size of several kilobytes or megabytes per image.
mechanism with heterogeneous traffic, but is too complex for Similarly, an acoustic sensor may capture data at a sample
real-time use. size of a few kilobits to megabits per second, depending on
Combining the MADDPG framework with NOMA, the sampling rate and resolution. In this work we assumed that
in [31], user association, power control and cache design the sensor nodes generate data for transmission in periodic
are jointly optimized to enhance energy efficiency in a intervals of time which is then transmitted to the access point.
dynamic terrestrial-satellite network. Yet, it has limitations on The access point first gathers the sample size and sampling
scalability and cannot adapt in real time for complex models interval as input parameters from sensor node through
involving satellites. The freshness-aware seed selection association request message. It applies proposed grouping
method presented in [32] is based on the assumption that stale algorithm to determine appropriate RAW and then assigns
content cannot satisfy cellular traffic offloading in OMNs, AID to the node or station based on selected RAW and update
and proposed a decay-based approach followed by a greedy the same through association reply.
one to facilitate efficient content propagation. However, While this approach is not standard for Wi-Fi, it allows
it does not perform well for dynamic user mobility adaptation the AP to optimize the assignment of the Restricted Access
and has high computational costs in large networks. Window (RAW) by applying our grouping algorithm using
The above Table 1 summarize the survey, outlining this additional data. The assignment of the Association ID
their grouping methods, simulation tools employed, and (AID) is deferred until after assessing these data properties
specific criteria utilized for grouping. The table highlights to enhance data transmission efficiency, thus optimizing
the algorithms used in comparative studies with which channel access based on specific traffic characteristics for
the evaluation has been done. Additionally, it lists the improved network performance. Additionally, this process
performance factors improved in the respective algorithms is conducted post-authentication of the node, addressing
such as mitigating interference (I), minimizing latency (L) or any security concerns associated with data exchange during
to enhance throughput (T). association. While our approach deviates from standard
The existing works discussed in the survey so far provides Wi-Fi HaLow mechanisms, it is designed to meet the unique
insight into the advancements progressed in the grouping requirements of our study.
methods for sensor network utilizing IEEE 802.11ah stan- The complete architecture of the proposed grouping
dard. The wide range of approaches has been designed to algorithm is described in Figure 4 which gets executed by
enhance the network performance factors but several crit- AP while associating the node. The entire process involves
ical gaps persist. Many approaches consider homogeneous the selection of RAW for assigning the node in an efficient
scenarios and incorporate static grouping mechanisms in way using the fuzzy logic approach named Takagi-Sugeno.
dynamic network conditions, which can lead to performance The initial stage starts by gathering information about the
degradation and lack adaptability. The goal of the novel node participating in association process through association
system is not only to address the limitations faced by request sent by them to the AP. Thus, sample interval and
existing grouping algorithms also develop an effective one sample size are extracted from the association request through
which adapt in real-time and to dynamic network conditions. the header field. The updated association request frame were
Our proposed algorithm’s motive is to design an intelligent shown in Figure 3. Sample interval is specified in terms of

185842 VOLUME 12, 2024

R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

seconds whereas Sample size is specified in bytes, and they

occupy 16 bits each in the header.

FIGURE 3. Updated association request frame.

A. COMPUTING TRANSMISSION EFFICIENCY RATE OF

INDIVIDUAL NODE (TERNI )
The transmission efficiency rate refers to the ratio of the
total RAW interval available for transmission compared to the
FIGURE 4. Proposed architecture of LBCA with FUZZY.
node’s sampling interval. It reflects the efficiency with which
the node can transmit the sampled data based on the total
available RAW interval. It mainly depends on the sampling 1) COMPUTING STANDARD DEVIATION OF RUR
interval. Higher rate implies the active participation in data The RAW utilization ratio refers to the sum of the RUR
transmission. As the beacon interval is divided into a fixed of individual nodes added into the respective RAW group.
number of RAW and the node will be assigned to one of them, Maintaining uniform channel utilization among all the groups
we compute the transmission efficiency rate with respect to is crucial for achieving load balancing thereby reducing the
the total RAW interval. collisions. The standard deviation of RUR among all the
Transmission Efficiency Rate TERNi is calculated using groups should always be the minimum for achieving high
Eq.(2) as the ratio of the total RAW interval (RI) to the sample uniformity. Initially the node’s RURNi is considered into RUR
interval (SI). Whenever the sampling interval is smaller than of the RAW group and standard deviation is calculated among
the RAW interval, the probability of node’s transmission the RAW groups by keeping RUR values of the remaining
efficiency will be 100%. groups unchanged. The standard deviation is calculated as
The Transmission Efficiency Rate depends on the RAW given in Eq.(5).
periodicity which is mentioned here as the total RAW interval s
and the ratio indicates the probability of node’s transmission P
(xi − µ)2
efficiency. This is one of the inputs= in the group selection σ = (5)
n
process.
1 (RI /SI ) > 1 Where
n
TERN i = (RI /SI ) (RI /SI ) < 1 (2) σ = Standard deviation
n = Number of fixed RAW
B. COMPUTING RAW UTILIZATION RATIO OF THE xi = Group i′ s RUR
INDIVIDUAL NODE (RURNI ) µ = Average RUR of all groups
RURNi refers to the probability that a node will be
transmitting or receiving data compared to the total available The pseudocode is represented below 1 to understand
time of the allocated RAW. The RAW utilization ratio of the about the process flow involved in calculation of standard
node depends on the sample size, supported data rate, and deviation in Eq.(5) which is the second input parameter for
sampling interval. RURNi is computed by multiplying the fuzzy architecture. This pseudocode gets executed at AP side
required transmission time of the node (RTNi ), TERNi and exactly at the block where estimation of the RAW’s standard
divided by RD as shown in Eq.(3) where RD is the duration deviation among RUR is mentioned in Figure 4.
of the individual RAW, whereas RTNi is estimated as the ratio The initializing step in line 1 indicates the start of the
of sample size to the supported data rate (SDR) as mentioned algorithm. The number of fixed RAW is declared as n in
in Eq.(4). line 2 which is used to represent how many groups have been
taken in the system, and the number of nodes taking part in
TERN i × RTN i simulation is declared as m in line 3. These parameters will
RURN i = (3)
RD be utilized for calculation and loop conditions in further steps.
SS The for loop begins in line 4 which gets executed whenever
RTN i = (4)
SDR the new node is being added to the RAW and then in line 5
the current RUR[j] value gets updated by adding the latest
C. INPUT PARAMETERS FOR GROUPING value of node’s RURNi estimated using the Eq. 3 and the loop
This section explains the computation of input parameters ends here. The procedure to estimate meanRUR is going to
required for the fuzzy logic. These parameters were computed begin which is indicated in line 6. Then the sum value S is
for each RAW in the channel separately. initialized as zero in line 7, and the for loop function starts

VOLUME 12, 2024 185843

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

Pseudocode for calculating SD provides an overview of the process carried out using fuzzy
1. Initialize logic to arrive at selecting the group with a higher probability.
2. n ← Number of fixed RAW The fuzzy logic is incorporated to handle the uncertainty
3. m ← Number of Nodes in the network and to make decisions intelligently by
4. for i ← 0 to m do: considering various input parameters. The major factors
5. RUR[j] ← RUR[j] + RURNi that need to be taken into account while performing group
6. Calculation of meanRUR allocation are traffic demand, channel resources, and effective
7. S=0 allocation mechanism. So, the inputs are taken as the number
8. for k ← 0 to n-1 do: of nodes which depicts the traffic, the standard deviation
9. sum=S+RUR[k] among RAW’s RUR is taken, which illustrates the deviation
10. meanRUR ← sum/n among RAW during utilization of the channel resource
11. Calculate variance and last one is the average transmission efficiency rate
12. sumsquareddiff ← 0 among the RAW in channel demonstrating that how well the
13. diff ← RUR[j] - meanRUR nodes were participating in transmission. The fuzzy logic
14. squareddiff ← diff * diff with all these parameters optimizes the grouping process
15. sumsquareddiff ← sumsquareddiff + squareddiff effectively by converting the crisp input into fuzzy values
16. variance ← sumsquareddiff / n then employs fuzzy rules to assess the situation and make
17. Calculate standardDeviation a decision about which RAW should be allocated based on
18. standardDeviation ← SquareRoot(variance) the input combinations and finally provides the grouping
probability by converting the fuzzy values into crisp values.
This approach allows refinement based on the demand and
from line 8 where the nodes get added among RAW in the enhances flexibility, thus improving network performance.
range of k to n-1. The sum used for calculating the mean is The fuzzy process flow involves four phases namely
updated by adding the RUR[k] with the S as mentioned in fuzzification, AND operation, Rule-based system, and
line 9. The meanRUR is then calculated as the average was defuzzification by centroid method which is described in
taken over the updated sum with the range executed in loop detail in the following sections.
condition as described in line 10. From line 11 the calculation
for variance will commence. The sumsquareddiff is set to
zero as in line 12 during the initial stage when the nodes are
not added. The difference between RUR[j] and meanRUR is
computed in line 13 as a term named diff, it gets squared as in
line 14 and mentioned as squareddiff . The squared value gets
added with the sum squared values to attain the updated value
of sumsquareddiff as in line 15 and the loop ends. Dividing
sumsquareddiff by the number of RAW yields the variance
that was mentioned in line 16. Then begins the computation of
standard deviation in line 17. Finally, the standard deviation
is evaluated as the square root of the variance which is given
in line 18 and the algorithm ends.
The block CHj mentioned in Figure 4 gets executed for
FIGURE 5. Fuzzy architecture.
each RAW whenever the node is sending association request
to AP. The number of nodes in the RAW is initially taken to
be zero and gets updated after the addition of each node in
the group. The first input parameter in fuzzy architecture is D. FUZZIFICATION
the number of nodes, second one is the standard deviation of The fuzzification is carried out using the membership
RUR among the groups which is estimated using algorithm function on the input parameters after being fed to the fuzzy
described above and the last one is the average TER of the network. The membership function for the number of nodes
RAW which is calculated using the transmission efficiency was given in Figure 6 with three categories to represent
rate of the individual nodes added from Eq.(6). the values, first the function is trapezoidal which indicates
P the range is small, and the second function is triangular for
TERN i a range of medium number of nodes and the last one is
ATER = (6)
n trapezoidal for representing higher range.
This procedure is repeated for all groups sequentially using The equations are detailed in Eqs.(7), (8), and (9) which
the CHj block by maintaining the values constant for groups break down the function µSmall , µMedium , andµHigh along
where the node is not added and changing the values for the with their corresponding limits. These functions are used for
group where it gets added. The fuzzy architecture in Figure 5 performing the fuzzification process to categorize the number

185844 VOLUME 12, 2024

R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

value being zero and values from low to high represents the
slope line from zero to one and value 1 to represent the high
value at the end of the function which maintains a straight
line at position 1. The triangular function has four constraints
which include the starting point of zero value and followed by
two slope lines one rising from zero to one and another one
descends back from one to zero and at last the point ending
at zero.

FIGURE 6. Membership function for number of nodes.

of nodes under these range of values.

µSmall


 1 0 ≤ x ≤ minn
 ((minn + maxn )/2) − x


minn < x

= ((maxn − minn )/2) FIGURE 7. Membership function for standard deviation of RUR.

≤ ((minn + maxn )/2)




The membership function for transmission efficiency rate


0 x ≥ ((minn + maxn )/2)
is showcased by two trapezoidal and a triangular function
(7) in Figure 8 with the three ranges of value: rare, moderate,
µMedium and frequent, their corresponding membership functions are
µRare , µModerate and µFrequent .


 0 x ≤ minn
x − minn


minn < x ≤ ((minn + maxn )/2)



= ((maxmaxn − minn )/2)
n−x

 ((minn + maxn )/2) < x ≤ maxn
((max − minn )/2)

n



0 x ≥ maxn


(8)
µHigh


 0 x ≤ ((minn + maxn )/2)
 x − ((minn + maxn )/2)


((minn + maxn )/2) < x

= ((maxn − minn )/2)
≤ maxn





1 FIGURE 8. Membership function for transmission efficiency rate.
x ≥ maxn
(9) The membership function of the grouping probability is
displayed in Figure 9 with the values spanning from min-
Where imum, normal, maximum, and extreme, by two trapezoidal
µ : MembershipValue and two triangular functions.
minn : MinimumNumberofthenodesamonggroups
max n : MaximumNumberofthenodesamonggroups

The membership function for the standard deviation of

RUR among the RAW groups was illustrated in Figure 7
with range being divided into slight, average, and greater,
the corresponding function values are µSlight , µAverage , and
µGreater , which is like the functions mentioned for number of
nodes. The first trapezoidal functions are defined with three
constraints: the value being one indicates a straight line at
1 position and the value from high to low describes the slope
line descending from 1 to 0 and the value is zero at the end
of the function. The last the trapezoidal function starts at a FIGURE 9. Membership function for grouping probabilities.

VOLUME 12, 2024 185845

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

TABLE 2. Fuzzy Inference Table.

E. RULE BASED SYSTEM TABLE 3. Simulation Parameter.

The rules are crafted based on the combination given in

inference Table 2, which illustrates about how the number of
nodes, the standard deviation of RAW utilization ratio and
average transmission efficiency rate among the RAW affects
the grouping probability. The results show that extreme
levels of grouping probability were obtained when there is
a smaller number of nodes with rare average transmission
efficiency rate and slight deviation among RAW utilization
ratio as mentioned in the rule R1, necessary conditions can
be outlined from the inference table for normal and minimum
levels of grouping probability as mentioned by R2 and R3.
R1: AVGTER(Rare) ∧ SDofRUR(Slight) ∧ NumberofNodes(Small) → GroupingProbability(Extreme)
R2: AVGTER(Moderate) ∧ SD of RUR(Average) ∧ Number of Nodes(Medium) → Grouping Probability
(Normal)
R3: AVGTER(Frequent) ∧ SD of RUR(Greater) ∧ Number of Nodes(High) → Grouping Probability
(Minimum)

F. AND OPERATION
The membership functions are utilized to transform the crisp IV. RESULTS AND DISCUSSION
input values into fuzzy values and those values are fed to The simulation parameter utilized in the network simulator
architecture for performing all possible combinations among is mentioned in Table 3 describes the duration and size
the inputs to formulate the rules R1, R2 and R3. The grouping of necessary parameters used during simulation. The plots
probability value is derived from the AND operation to arrive mentioned below are generated from the simulation in NS-
at the minimum value among the inputs which is explained 3 using the proposed LBCA grouping scheme.
below by taking one of the rules from the inference table. In our simulation, traffic intensity varies as follows: 30 to
RULE: AVGTER(Frequent) ∧ SD of RUR(Slight) ∧ Number of Nodes(Small) → Grouping Probability
70 bytes for approximately 60% of the nodes, 100 to
AND:
(Maximum)
0.7 ∧ 0.5 ∧ 0.4 → 0.4
200 bytes for 30% of the nodes, and 2000 to 4000 bytes
for 10% of the nodes, representing the characteristics of a
G. DEFUZZIFICATION CENTROID METHOD heterogeneous network.
The Channel Utilization Ratio of a RAW Group refers
x × µg (x)dx
R
CentroidBackoff = (10) to the proportion of the available channel time during a
µg (x)
R
Restricted Access Window (RAW) period that is successfully
The output is attained by applying the centroid method used by a group of devices for data transmission. It measures
of defuzzification on R1, R2 and R3 as outlined in Eq.(10). how efficiently the RAW group’s allocated time in the
We evaluated the rules for each value to determine what channel were being utilized.
were their degree of membership. These membership values We compare our results over random method to analyse
are then utilized through the defuzzification process with and prove that how our method achieves better performance.
the centroid method to obtain a crisp output. x means that The random method assigns nodes to RAW groups without
crisp output value whose membership to fuzzy output will considering specific network characteristics, leading to load
be equal 1, and µg (x) is the membership value of the fuzzy imbalances, particularly in heterogeneous networks where
output evaluated by the rule. Once this process is over sample sizes and sampling intervals vary among nodes.
the group with the best grouping probability is selected This imbalance can cause increased collisions and delays.
and the node is added to that group. Subsequently, their While it is possible to adjust the number of RAW groups,
corresponding CUR and CAR are updated along with the AID their duration, and the number of slots, achieving optimal
allocation through association reply. Thus, we have discussed performance with the random method would require a
the proposed LBCA grouping mechanism using Fuzzy logic more strategic approach rather than random allocation. Our
in this section detailing the exact conditions required for work aims to address these challenges by implementing an
efficient utilization of the channel. organized grouping strategy to enhance performance.

185846 VOLUME 12, 2024

R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

The results visually depicted in Figures 10 and 11 illustrate

the average channel utilization rate for five fixed RAW over
time comparing the proposed LBCA and random grouping
scheme from [25].
Both graphs exhibit a constant increase in average CUR
over time since when the nodes get added to the group
the CUR increases as the utilization of channel resource
increases. The LBCA graph showcases consistent and
balanced utilization of channel among the groups till the
end of the simulation. In contrast, in a random method, it is
not consistent at all due to the random nature of grouping.
Also, performance among groups varies as the distribution of
channel resource is unfair. Some groups reach high utilization
faster and others lagging.
Additionally, it is evident that all the groups are reaching
the saturation state nearly around 370-386 seconds, but FIGURE 11. Average channel utilization ratio of nodes - random method.
observing the random method shows that the saturation
state is attained at different times which is broadcasting the
larger deviation happening in channel utilization between the method, as the graph of the random method shows that the
groups. The graphs also describe the time taken in LBCA collision keeps on increasing, lacking uniformity among the
method is almost around 386 seconds whereas in random groups while in LBCA all the groups have more or less
method for the same number of nodes, it takes around the same amount of collision and also the collision rate is
482 seconds. Thus, LBCA provides significant improvement constant not having much deviation thus mitigating collision
in channel utilization in terms of reduction in time and due to the uniform distribution of nodes among the groups.
close clustering among groups indicates effective channel
utilization by the nodes.

FIGURE 12. Average collision rate of nodes - LBCA method.

FIGURE 10. Average channel utilization ratio of nodes - LBCA method.

The drop in throughput is due to congestion, increased
The collision occurs when the node is not able to perform collision, or interference over time. The rate of decrease
its transmission due to several factors. The number of in throughput values across the groups varies, some have
collisions and collision rate increases when the node does not sharper drops whereas the remaining group doesn’t have that
get channel access at an appropriate time. From the graph, much drop indicating certain kinds of interference hindering
we can recognize that LBCA outperforms the random method the transmission.
in terms of collision rate and number of collisions among the The plots in Figures 14 and 15 show significant improve-
groups. ment in throughput of LBCA over random method here also
The plots visualized in Figures 12 and 13 details the we can see that the random method throughput is dropping
collision rate of the proposed and random method from which from 1Mbps to around 800kbps for groups 4 and 5 and about
we can notice that the collision rate in LBCA is around 850kbps for the other three groups whereas the throughput
0.11 but in random it is around 0.17, additionally the number is maintained constant in the LBCA method of around
of collisions is more in the random method than the LBCA 0.9Mbps in all groups. As the nodes keep on joining the

VOLUME 12, 2024 185847

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

recording the time just before sending the DATA frame and
again after receiving the ACK at the MAC layer.
The plots depicted in Figures 16, 17 explain the delay
in LBCA and Random method over beacon intervals. The
LBCA has a very low initial delay which is nearly at 0 for
all groups but in the random method the initial delay is high
and the rate of increase over beacon interval is faster. The
LBCA approach has a steady increase indicating that the load
balance across groups is done effectively to manage traffic
and reduce significant fluctuation in delay. Thus, LBCA
provides better control over the delay, ensures predictable
increase among all the groups, and also has a lower delay
of values just above 0.005 when compared with the random
method having a delay of around 0.007.

FIGURE 13. Average collision rate of nodes - random method.

group periodically, the throughput start to decline due to the

transmission of a larger amount of data.

FIGURE 16. Average delay of LBCA method.

FIGURE 14. Individual node’s throughput of LBCA method.

FIGURE 17. Average delay of random method.

The plots are shown in Figures 18, 19, and 20 present

the average channel utilization rate (CUR), collision rate
FIGURE 15. Individual node’s throughput of random method. (CR) and throughput (TR) of both methods. Notably in
average CUR plot shows that the average channel utilization
Delay in our study is defined as the sum of transmission, of the LBCA method is around 0.9 to 0.8 for all the
contention, and propagation delays. It is calculated by groups whereas in the random method, groups 2 and 4 have

185848 VOLUME 12, 2024

R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

values around 0.9 but for all the three other groups it
fluctuates between 0.6 to 0.7 which indicates the load is
not distributed evenly but in LBCA method it maintains
constant values in all groups indicating efficiency and load
balance.
The minimal variation in the average value among groups
indicates consistent utilization of the channel by all the
groups, similarly high values of overall channel utilization
across all the groups indicate balanced usage of the available
resource. This proves LBCA method is more efficient than the
random having more variation among groups and a medium FIGURE 20. A) RAW’s Average Throughput of LBCA Method B) RAW’s
overall value. Average Throughput of Random Method.

V. CONCLUSION
The proposed LBCA algorithm, with an intelligent group-
ing approach using fuzzy logic and a standard deviation
strategy for optimizing different types of parameters in the
network, enhances the performance by improving multiple
key metrics. LBCA also maintains an almost perfect balance
on the distribution of loads among RAW, making significant
refinements compared to the random method. The result
shows a 1.37% increase in throughput, ensuring all sensor
nodes transmit maximum data over the channel due to the
FIGURE 18. A) RAW’s Average Channel Utilization Ratio of LBCA Method appropriate access mechanism. The channel utilization rate
B) RAW’s Average Channel Utilization Ratio of Random Method. increases by 7.65%, suggesting more effective load balancing
and stable high-level channel usage.
The plots in Figure 19 illustrate the average collision rate This algorithm effectively manages to minimize the
in two methods and gives clear cut view that the LBCA collided packet. The results show a promising reduction
method is having a value between 0.05 to 0.06 resulting of 25% in the number of collisions and a decrease of
in a total number of 162 collisions whereas in random it 16.42% in collision rate (rate of successful transmission
is from 0.065 to 0.068 with the total number of collisions without interference) among groups. In addition, the delay
being 216 clearly indicating that proposed LBCA outperform is 8.57% lower than the random method ensuring faster and
the random method in terms of collision due to the proper reliable data transmission. The results indicate that the LBCA
utilization of grouping strategy in allocation of the nodes approach improves throughput and channel utilization and
among RAW. minimizes delay and collisions.
In future work, we will concentrate on optimizing the
RAW allocation process in saturated cases, incorporating
ML techniques for more accurate prediction of the channel
duration to improve network efficiency and will further focus
on the slot division according to dynamic traffic conditions.

REFERENCES
[1] W. Yin, P. Hu, W. Wang, J. Wen, and H. Zhou, ‘‘FASUS: A fast association
mechanism for 802.11ah networks,’’ Comput. Netw., vol. 175, Jul. 2020,
Art. no. 107287, doi: 10.1016/j.comnet.2020.107287.
[2] L. Tian, S. Santi, A. Seferagić, J. Lan, and J. Famaey, ‘‘Wi-Fi HaLow
for the Internet of Things: An up-to-date survey on IEEE 802.11ah
research,’’ J. Netw. Comput. Appl., vol. 182, May 2021, Art. no. 103036,
FIGURE 19. A) RAW’s Average Collision Rate of LBCA Method B) RAW’s doi: 10.1016/j.jnca.2021.103036.
Average Collision Rate of Random Method. [3] Y. J. Wu, T. L. Sheu, and P. H. Chan, ‘‘Dynamic slot allocation in restricted
access window for IEEE 802.11 ah networks,’’ J. Inf. Sci. Eng., vol. 38,
The plot visualization in Figure 20 describes the average pp. 697–708, May 2022, doi: 10.6688/JISE.202105_37(3).0012.
throughput of LBCA and random method, it is observed [4] L. Tian, S. Santi, S. Latré, and J. Famaey, ‘‘Accurate sensor traffic
estimation for station grouping in highly dense IEEE 802.11ah networks,’’
that the throughput of the LBCA scheme is approximately in Proc. 1st ACM Int. Workshop Eng. Reliable, Robust, Secure Embedded
700kbps but for random it is around 610kbps thus the LBCA Wireless Sens. Syst., Nov. 2017, pp. 1–9, doi: 10.1145/3143337.3149819.
outperforms the random method in terms of throughput [5] E. Zazhigina, R. Yusupov, A. Lyakhov, and E. Khorov, ‘‘Analytical study of
restricted access window with short slots for fast and reliable data delivery
due to effective channel utilization and achieving minimum from energy-harvesting sensors,’’ Comput. Netw., vol. 250, Aug. 2024,
collision. Art. no. 110573, doi: 10.1016/j.comnet.2024.110573.

VOLUME 12, 2024 185849

 R. Nandhini, R. Radha: Fuzzy Logic Driven Load Balanced Channel Allocation Algorithm

[6] S. Aust. (2011). Proposed IEEE 802.11ah Use Cases. [Online]. Available: [25] M. Mahesh and V. P. Harigovindan, ‘‘Hidden terminal aware grouping
https://mentor.ieee.org/802.11/dcn/11/11-11-0017-00-00ah-proposed- scheme for IEEE 802.11ah based dense IoT networks,’’ Comput. Commun.,
ieee-802-11ah-use-cases.pdf vol. 191, pp. 161–172, Jul. 2022, doi: 10.1016/j.comcom.2022.04.033.
[7] T. Adame, A. Bel, B. Bellalta, J. Barcelo, and M. Oliver, ‘‘IEEE 802.11AH: [26] L. Tian, M. T. Mehari, S. Santi, S. Latré, E. De Poorter, and J. Famaey,
The WiFi approach for M2M communications,’’ IEEE Wireless Commun., ‘‘Multi-objective surrogate modeling for real-time energy-efficient station
vol. 21, no. 6, pp. 144–152, Dec. 2014. grouping in IEEE 802.11ah,’’ Pervas. Mobile Comput., vol. 57, pp. 33–48,
[8] F. Malook, O. Mujahid, Z. Ullah, and T. Fouzder, ‘‘On enhancing the Jul. 2019, doi: 10.1016/j.pmcj.2019.04.007.
performance of IEEE 802.11ah by employing a dynamic raw approach in [27] I. Surenther, K. P. Sridhar, and M. K. Roberts, ‘‘Maximizing energy
IoT networks,’’ Wireless Pers. Commun., vol. 129, no. 3, pp. 1983–1997, efficiency in wireless sensor networks for data transmission: A deep
Apr. 2023, doi: 10.1007/s11277-023-10221-w. learning-based grouping model approach,’’ Alexandria Eng. J., vol. 83,
[9] M. A. Mondal and M. I. Hussain, ‘‘Station grouping mechanism using pp. 53–65, Nov. 2023, doi: 10.1016/j.aej.2023.10.016.
machine learning approach for IEEE 802.11ah,’’ Ad Hoc Netw., vol. 149, [28] P. Sthapit and J.-Y. Pyun, ‘‘Station grouping strategy for minimizing
Oct. 2023, Art. no. 103238, doi: 10.1016/j.adhoc.2023.103238. association delay in IEEE 802.11ah,’’ IEICE Trans. Commun., vol. 100,
no. 8, pp. 1419–1427, 2017, doi: 10.1587/transcom.2016ebp3306.
[10] C.-M. Huang and C.-H. Hsieh, ‘‘Registered-backoff-time (RBT) based
[29] T.-C. Chang, C.-H. Lin, K. C. Lin, and W.-T. Chen, ‘‘Load-balanced sensor
channel access with grouping control for the trigger RAW mode of
grouping for IEEE 802.11ah networks,’’ in Proc. IEEE Global Commun.
IEEE 802.11ah IoT network,’’ Comput. Netw., vol. 242, Apr. 2024,
Conf. (GLOBECOM), Dec. 2015, pp. 1–6.
Art. no. 110209, doi: 10.1016/j.comnet.2024.110209.
[30] E. Garcia-Villegas, A. Lopez-Garcia, and E. Lopez-Aguilera, ‘‘Genetic
[11] L. Tian, E. Khorov, S. Latré, and J. Famaey, ‘‘Real-time station grouping
algorithm-based grouping strategy for IEEE 802.11ah networks,’’ Sensors,
under dynamic traffic for IEEE 802.11ah,’’ Sensors, vol. 17, no. 7, p. 1559,
vol. 23, no. 2, p. 862, Jan. 2023, doi: 10.3390/s23020862.
Jul. 2017, doi: 10.3390/s17071559.
[31] X. Li, H. Zhang, H. Zhou, N. Wang, K. Long, S. Al-Rubaye, and
[12] B. T. Guedes, D. Passos, and F. G. O. Passos, ‘‘A spectral G. K. Karagiannidis, ‘‘Multi-agent DRL for resource allocation and cache
clustering algorithm for intelligent grouping in dense wireless design in terrestrial-satellite networks,’’ IEEE Trans. Wireless Commun.,
networks,’’ Comput. Commun., vol. 198, pp. 117–127, Jan. 2023, vol. 22, no. 8, p. 1, Aug. 2023, doi: 10.1109/TWC.2022.3231379.
doi: 10.1016/j.comcom.2022.11.017. [32] H. Zhou, X. Chen, S. He, C. Zhu, and V. C. M. Leung, ‘‘Freshness-aware
[13] U. Sangeetha and A. V. Babu, ‘‘Service differentiation in IEEE seed selection for offloading cellular traffic through opportunistic mobile
802.11ah WLAN under restricted access window based MAC pro- networks,’’ IEEE Trans. Wireless Commun., vol. 19, no. 4, pp. 2658–2669,
tocol,’’ Comput. Commun., vol. 172, pp. 142–154, Apr. 2021, doi: Apr. 2020, doi: 10.1109/TWC.2020.2967658.
10.1016/j.comcom.2021.03.017.
[14] M. Mahesh, B. S. Pavan, and V. P. Harigovindan, ‘‘Data rate-based
grouping using machine learning to improve the aggregate through-
put of IEEE 802.11ah multi-rate IoT networks,’’ in Proc. IEEE Int.
Conf. Adv. Netw. Telecommun. Syst. (ANTS), Dec. 2020, pp. 1–5, doi:
10.1109/ANTS50601.2020.9342758.
[15] C. Kai, J. Zhang, X. Zhang, and W. Huang, ‘‘Energy-efficient sen-
sor grouping for IEEE 802.11ah networks with max-min fairness
guarantees,’’ IEEE Access, vol. 7, pp. 102284–102294, 2019, doi:
10.1109/ACCESS.2019.2931709. R. NANDHINI received the Graduate degree from
[16] T.-C. Chang, C.-H. Lin, K. C. Lin, and W.-T. Chen, ‘‘Traffic-aware the SSN College of Engineering, the B.E. degree in
sensor grouping for IEEE 802.11ah networks: Regression based analysis
electronics and communication engineering from
and design,’’ IEEE Trans. Mobile Comput., vol. 18, no. 3, pp. 674–687,
the A. V. C. College of Engineering, in 2012,
Mar. 2019, doi: 10.1109/TMC.2018.2840692.
and the M.E. degree in VLSI design, in 2015.
[17] S. Bhandari, S. K. Sharma, and X. Wang, ‘‘Device grouping for fast
She is currently pursuing the Ph.D. degree with
and efficient channel access in IEEE 802.11ah based IoT networks,’’ in
Proc. IEEE Int. Conf. Commun. Workshops (ICC Workshops), May 2018, Vellore Institute of Technology, Chennai. She was
pp. 1–6. an Assistant Professor with the C. K. College of
[18] M. Mahesh, B. S. Pavan, and V. P. Harigovindan, ‘‘Data rate-based Engineering, in 2016. She has done research in
grouping to resolve performance anomaly of multi-rate IEEE 802.11ah IoT MEMS-based piezoelectric cantilevers for energy
networks,’’ IEEE Netw. Lett., vol. 2, no. 4, pp. 166–170, Dec. 2020, doi: harvesting using COMSOL multiphysics software and IntelliSuite. She
10.1109/LNET.2020.2998469. recently working with IoT communication protocols and MEMS sensors.
[19] R. Nishida, M. Shimokawa, K. Sanada, H. Hatano, and K. Mori,
‘‘A station grouping method for IEEE 802.11ah networks with various
types of network non-uniformity,’’ Nonlinear Theory Its Appl., IEICE,
vol. 14, no. 1, pp. 35–49, 2023, doi: 10.1587/nolta.14.35.
[20] S.-G. Yoon, J.-O. Seo, and S. Bahk, ‘‘Regrouping algorithm to alleviate
the hidden node problem in 802.11ah networks,’’ Comput. Netw., vol. 105,
pp. 22–32, Aug. 2016, doi: 10.1016/j.comnet.2016.05.011.
[21] W. Damayanti, S. Kim, and J.-H. Yun, ‘‘Collision chain mitiga-
tion and hidden device-aware grouping in large-scale IEEE 802.11ah
R. RADHA received the Ph.D. degree from Anna
networks,’’ Comput. Netw., vol. 108, pp. 296–306, Oct. 2016, doi:
10.1016/j.comnet.2016.09.006. University, Chennai, in 2016. Her Ph.D. thesis
titled ‘‘Collision avoidance strategies over multi
[22] H. Mosavat-Jahromi, Y. Li, and L. Cai, ‘‘A throughput fairness-based
grouping strategy for dense IEEE 802.11ah networks,’’ in Proc. IEEE hop wireless ad hoc networks.’’ She is currently
30th Annu. Int. Symp. Pers., Indoor Mobile Radio Commun. (PIMRC), an Associate Professor with the School of Com-
Sep. 2019, pp. 1–6. puter Science and Engineering, Vellore Institute
[23] L. Zhang, H. Li, Z. Guo, L. Ding, F. Yang, and L. Qian, ‘‘Signal of Technology, Chennai. She is also an Anna
strength assistant grouping for lower hidden node collision probability University Rank holder during her PG. She has
in 802.11ah,’’ in Proc. 9th Int. Conf. Wireless Commun. Signal Process. completed two research projects funded by AICTE
(WCSP), Oct. 2017, pp. 1–6. and TNSCST. She has published multiple research
[24] C.-C. Hu, Z.-B. Liu, Y.-M. Lin, and G.-J. Xu, ‘‘An efficient and effective papers in various SCI, Scopus indexed journals, and conferences. Her
regrouping algorithm for minimizing hidden pairs in 802.11ah networks,’’ research interests include mobile ad-hoc networks, the IoT communication
in Proc. Int. Conf. Adv. Comput. Technol., Inf. Sci. Commun., 2019, technologies, cross layer deign, and the IoT for smart cities.
pp. 191–196, doi: 10.5220/0008097301910196.

185850 VOLUME 12, 2024