Chinese Journal of Electronics

Vol.32, No.4, July 2023

Enhancing Network Throughput via the Equal

Interval Frame Aggregation Scheme for
IEEE 802.11ax WLANs
ZHU Yihua and XU Mengying
(School of Computer Science and Technology, Zhejiang University of Technology, Hangzhou 310023, China)

Abstract — Frame aggregation is fully supported in power, data rates applied at nodes in the WSNs are
the newly published IEEE 802.11ax standard to improve
quite low, usually having tens of Kbps. In order to de-
throughput. With frame aggregation, a mobile station
combines multiple subframes into an aggregate MAC ser-
liver data in a higher rate, some WSN applications ac-
vice data unit (A-MSDU) or an aggregate MAC protocol commodate Wi-Fi technology. The examples include:
data unit (A-MPDU) for transmission. It is challenging Wi-Fi enabled sensors were introduced [3]; Wi-Fi chip-
for a mobile station in 802.11ax WLANs to set an appro- sets were applied in monitoring temperature, humidity,
priate number of subframes being included in an A-
volatile organic compounds and miniature dust particles
MSDU or A-MPDU. This problem is solved in this paper
by the proposed equal interval frame aggregation (EIFA) [4]; Wi-Fi embedded network processors were used in
scheme which lets a mobile station aggregate at most k monitoring heart rate [5]; and Wi-Fi module was used
subframes at a fixed time period of T. A novel Markov in the car that can sense obstruction, temperature, gas,
model is developed for deriving the probability of num- fire, etc. [6]. In a word, the IEEE 802.11 standard is fre-
ber of subframes in the data buffer at the mobile station,
quently applied in the WSN applications.
resulting in the throughput and packet delay in the EIFA.
Moreover, the optimization problem of maximizing the In fact, the IEEE 802.11 family of standards has
throughput with the constraint on delay is formulated, acted as one of the most popular technologies for wire-
and its solution leads to the optimal pair of parameters k less connectivity. Nodes in the IEEE 802.11 standard
and T for improving throughput in the EIFA scheme. based WLAN are termed as mobile stations (STAs).
Simulation results show the EIFA has a higher through-
Since the birth of the first IEEE 802.11 standard [1] is-
put than the ones in which the mobile station chooses the
minimum, the maximum, or a random number of sub- sued in 1997, members of the 802.11 family are increas-
frames. ing. The newest member is the IEEE 802.11ax stand-
Key words — IEEE 802.11ax standard, Frame ag- ard [7], released in May 2021, called Wi-Fi 6, which in-
gregation, WLAN, Throughput, Markov model, MAC.
troduces some new features, including the orthogonal
frequency division multiple access (OFDMA) that sup-
ports the STAs to simultaneously transmit over mul-
tiple resource units (RUs) [8].
I. Introduction
A MAC frame, namely MAC protocol data unit
The Internet of things (IoT) plays a key role in col- (MPDU), carries MAC service data unit (MSDU). More
lecting data from the environment. Wireless local area specifically, MPDU consists of MSDU and the MAC
networks (WLANs) and wireless sensor networks header and the MAC tail. Here, the header and the tail
(WSNs) are main components of the IoT. Typically, are overhead. Clearly, in the case when the size of the
WLAN and WSN adopt IEEE 802.11 standard [1] and MAC header and tail is close to or greater than that of
IEEE 802.15.4 standard [2] in the medium access con- MSDU, efficiency of channel usage is low.
trol (MAC) and the physical layers, respectively. The Apart from the MAC header and tail, channel con-
popular Wi-Fi is based on the IEEE 802.11 standard. tention (e.g., backoff) and acknowledgement (ACK) of
Due to IEEE 802.15.4 standard targeting at low frame also yield overhead. To reduce overhead, frame

Manuscript Received Aug. 21, 2022; Accepted Jan. 6, 2023. This work was supported by the National Natural Science Foundation
of China (61772470, 61432015) and the National Key R&D Program of China (2019YFD0901605).
© 2023 Chinese Institute of Electronics. DOI:10.23919/cje.2022.00.282
748 Chinese Journal of Electronics 2023

aggregation is introduced in IEEE 802.11n standard [9] This paper has the contributions as follows:
in 2009 and extended by the newest Wi-Fi 6 in 2021 [7]. • We propose the equal interval frame aggrega-
With frame aggregation, a STA encapsulates multiple tion (EIFA) scheme that aims to improve throughput.
short frames into one longer frame so that the multiple In the EIFA, there are two parameters k and T . With
short frames share the header and tail of the MAC the EIFA scheme, each STA aggregates at most k sub-
frame, the preamble and header of the physical layer, frames in its data buffer at a fixed time period of T to
the time consumed in channel contention, and ACK generate a longer frame. It should be stressed that the
frame. It was reported that efficiency of channel usage subframes in the EIFA scheme can be with different
and network throughput can be considerably improved TIDs. This characteristic differs from the existing frame
when the STAs adopt frame aggregation. aggregation schemes based on the IEEE 802.11 stand-
Before the birth of Wi-Fi 6, the MPDUs must all ards prior to Wi-Fi 6.
be of the same 802.11e quality of service (QoS) access • We develop a novel Markov chain to model the
category when A-MPDU frame aggregation is used, and EIFA, from which we derive the probability of the
voice MPDUs cannot be mixed with best effort or video number of the subframes in the data buffer at the STA.
MPDUs within the same aggregated frame. Wi-Fi 6, Then, we derive the throughput and delay in the EIFA.
however, introduces multi-traffic identifier aggregated In addition, we formulate the optimization problem
MPDU (Multi-TID A-MPDU) that supports the ag- that maximizes the throughput with respect to the pair
gregation of frames from multiple traffic identifiers of parameters k and T under delay constraint. The
(TIDs). It is Wi-Fi 6 that has the capability of mixing solution to the optimization problem leads to the op-
MPDUs, i.e., it can integrate frames from the same or timal values of the parameters k and T , which are ap-
different QoS access categories. The ability of combin- plied in the EIFA scheme.
ing MPDUs from different QoS traffic classes allows • Simulation results show that the proposed EIFA
802.11ax radios to aggregate more efficiently by trans- achieves a higher throughput than the ones in which a
mitting over multiple RUs, thus reducing overhead and STA chooses the minimum NoS, the maximum NoS, or
increasing overall network efficiency. a random NoS.
We notice that the existing frame aggregation The rest of the paper is organized as follows. The
schemes have one or more of the following shortcom- related researches are surveyed in Section II. We out-
ings. First, they do not take into account the OFDMA line frame aggregation and present the EIFA with its
mechanism, a new feature introduced in the newest Markov model in Section III, formulate the optimiza-
IEEE 802.11ax standard, which supports simultaneous tion problem in Section IV, evaluate the performance of
transmission over multiple RUs. the EIFA via simulation in Section V, and conclude the
Second, the IEEE 802.11ax standard only defines paper in Section VI.
the range of the number of subframes (NoS), but the
STAs do not have the knowledge of how many sub- II. Related Work
frames should be included in generating a longer frame. As previously mentioned, the OFDMA mechanism
Here, subframes are the short frames being aggregated. is one of the new features in the newly-published IEEE
Third, in the case when traffic to a STA is intermittent, 802.11ax standard, which allows access point (AP) to
the STA needs to wait for multiple subframes to arrive divide wireless channel into smaller chunks (i.e., RUs)
and then aggregate them to produce a longer frame. and allocate them to different users. The OFDMA
Obviously, a shorter waiting time may result in only a mechanism has attracted much interest in research
few subframes being aggregated, which may decrease ef- community. Kotagiri et al. [10] proposed a distributed
ficiency of channel usage, whereas a longer waiting time RU selection method using convolutional neural net-
causes a longer packet delay although it helps in ag- work based deep reinforcement learning to improve
gregating more subframes to improve efficiency of chan- throughput and latency. Xie et al. [11] developed a
nel usage. Obviously, it is time-consuming in practice multi-dimensional busy-tone arbitration mechanism to
for a STA to test all the NoSs with different TIDs for reduce collisions among STAs contending for the same
the best NoS. RU. Zheng et al. [12] studied the OFDMA-based ran-
In summary, it is urgent to design a frame aggrega- dom access backoff and presented the retransmission
tion scheme that both takes into account the OFDMA number aware channel access (RNACA) scheme for
mechanism and supports the STAs in the 802.11ax IEEE 802.11ax-based WLAN, which lets the STA in-
WLAN to find the optimal pair of NoS and waiting tending to retransmit a frame make decision on wheth-
time so that efficiency of channel usage is enhanced. er to double its OFDMA contention window size or not
This is the motivation of our work. by using a probability resulting from the number of re-
 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 749

transmissions, number of RUs, and number of STAs. spatial multiplexing, and then computed the through-
To avoid collision in the uplink OFDMA random ac- put and the overhead in the 802.11ac network. Saldana
cess (UORA), Lee et al. [13] proposed the method with et al. [27] explored the throughput versus latency
non-orthogonal multiple access (NOMA). Lanante et al. tradeoff in the context of central controlled solutions.
[14] equipped UORA with carrier sensing capability and Saldana et al. [28] proposed an algorithm for the dy-
developed the hybrid UORA OFDMA access mechan- namic tuning of the maximum size of aggregated frames
ism to reduce collisions. Cheng et al. [15] presented an in 802.11 WLANs, which allows the network manager
analytical model to investigate the performance of to find the optimal balance between throughput and
UORA in transient condition under burst arrivals with latency.
arbitrary distribution, which leads to access success Abedi et al. [29] focused on finding the optimal
probability, average access delay, cumulative distribu- length of an A-MPDU and developed a standard com-
tion function (CDF) for the number of transmissions, pliant, practical, near-optimal frame aggregation al-
and utilization of UORA. In addition, Ghanem et al. [16] gorithm (PNOFA) in order to maximize throughput.
considered resource allocation for multi-user downlink Morimoto et al. [30] analyzed throughput and delay in
multiple input single output OFDMA system. And full-duplex WLANs with frame aggregation. Liu et al.
Khoramnejad et al. [17] investigated load management, [31] proposed an adaptive frame length aggregation
power and admission control in downlink cellular OF- strategy with remaining delay limit in VANETs. Kim et
DMA networks. al. [32] analyzed how different overhead components im-
Although frame aggregation is designed for 802.11 pact the efficiency of multi-user transmission and
WLAN, it has been applied in WSN. Xiao et al. [18] presented a scheme for constructing a multi-user frame
proposed an opportunistic data aggregation algorithm with the optimal length to maximize transmission effi-
to support the data collection in low-duty-cycle WSNs ciency. Liu et al. [33] optimized the number of the ag-
with unreliable links. gregated sub-MAC protocol data units to maximize the
Lin et al. [19] proposed an analytical model to throughput of the two-level frame aggregation accord-
study the performance of the MAC protocol by using ing to the subframe size, the maximum aggregation
the two frame aggregation techniques, namely MPDU level and the real-time channel bit-error-rate (BER).
aggregation (A-MPDU) and MSDU aggregation (A- Suzuki et al. [34] constructed an optimization problem
MSDU), under uni-directional and bi-directional data that maximizes theoretical throughput considering bit
transfer. Kowsar et al. [20] analyzed performance of error in A-MSDU under the IEEE 802.11ac standard
combined A-MSDU and A-MPDU via NS-3 simulation. and proposed a scheme to determine the optimal sub-
Abdallah et al. [21] developed a framework for jointly frame sets through an exhaustive search for both A-
selecting the data rate, the multiple input multiple out- MPDU and A-MSDU.
put (MIMO) mode and the frame aggregation configur- Kim et al. [35] proposed an analytical model based
ation at the AP using subcarrier-level channel state in- on an enhanced discrete time Markov chain to describe
formation. Assasa et al. [22] considered impact of the the post-backoff behavior due to frame aggregation, and
traffic patterns in the wireless network on frame aggreg- evaluated the throughput performance. Seytnazarov et
ation, and then they presented a waiting policy for the al. [36] proposed a Markov chain model to estimate the
uplink case that either waits for a minimum number of average aggregation size resulting from block acknow-
packets or for a maximum amount of time, whichever ledgement window operations under noisy channel con-
comes first. ditions, and the obtained average aggregation size was
Nomura et al. [23] proposed an efficient frame ag- used in enhancing the performance of the aggregation-
gregation scheme that improves system throughput and enabled WLANs. Jabri et al. [37] adopted Markov mod-
decreases frame error rate for downlink multi-user el for 802.11n MAC layer to remedy the drawback in
MIMO transmissions in IEEE 802.11ac WLANs. the available analytical models in which the impact of
Charfi et al. [24] proposed a scheduling algorithm the block ACK window limit on the aggregated frame
based on urgency delay and an adaptive frame aggrega- size is ignored. Lu et al. [38] presented the timely data
tion technique to obtain an optimal trade-off between delivery (TDD) scheme and a Markov chain model for
maximizing throughput and minimizing delay. the TDD to reduce packet delay for energy-harvesting
Zhou et al. [25] proposed an adaptive frame length IoT devices.
aggregation scheme for vehicular ad hoc networks The EIFA scheme proposed in this paper differs
(VANETs), aiming at improving transmission effi- from the existing schemes in that: 1) NoS and waiting
ciency and increasing data throughput. Yazid et al. [26] time are jointly optimized, which yields the optimal
modeled the 802.11ac station enabling A-MPDU and pair of the parameters to maximize network through-
750 Chinese Journal of Electronics 2023

put; and 2) The OFDMA mechanism introduced in the IEEE 802.11ax standard limits the maximum num-
newest IEEE 802.11ax standard is considered. ber of MPDU subframes to 64 [7]. It, however, leaves
the optimal NoS unsolved. This brings a challenge to
III. The EIFA Scheme with Mathemat- the STAs that need suitable NoS in frame aggregation.
ics Model We will address the problem via a Markov model in the
Frame aggregation is classified by one-level and next section.
two-level. The one-level frame aggregation consists of This paper focuses on one-level frame aggregation.
A-MSDU and A-MPDU, shown in Fig.1(a) and (b), re- 1. The EIFA Scheme
spectively [39], where DA and SA stand for destination In this paper, we study the WLAN in which all
address and source address, respectively. The two-level STAs adopt IEEE 802.11ax standard in their MAC and
frame aggregation is shown in Fig.1(c), which combines physical layers and the STAs apply the proposed EIFA
A-MSDU and A-MPDU. scheme in delivering data.

PSDU

PHY header MAC header A-MSDU FCS

Subframe #1 Subframe #2 Subframe #3 ...

DA SA Length MSDU Padding

Octets: 6 6 2 Variable 0-3

(a) A-MSDU

PSDU

PHY header A-MPDU

Subframe #1 Subframe #2 Subframe #3 ...

MPDU delimiter

Reserved MPDU length CRC Signature MPDU Padding

Bits: 4 12 8 8

Octets: 4 Variable 0-3

(b) A-MPDU

PSDU (A-MPDU)

PHY header MPDU subframe #1 MPDU subframe #2 ...

MPDU delimiter MPDU Padding

MAC header MSDU subframe #1 MSDU subframe #2 ... FCS

A-MSDU
(c) Two-level frame aggregation

Fig. 1. Two levels of frame aggregation.

 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 751

In the sequel, we refer to the frame generated by completion of frame aggregation; and
aggregating subframes as a “long frame.” Accordingly, 4) Only one long frame is generated and transmit-
the subframes being aggregated are termed as “short ted in an aggregation period.
frames.” The EIFA scheme is illustrated in Fig.2, where the
Our EIFA scheme is characterized by the follow- proposed EIFA scheme resides in the MAC layer in the
ing features: open system interconnection (OSI) reference model. The
1) At every fixed time interval of T , called aggreg- EIFA uses the data buffer to keep the traffic from the
ation period below, a STA in the WLAN aggregates at upper layer, and it aggregates at most k short frames to
most k short frames to generate one long frame; generate one long frame at every aggregation period.
2) Frame aggregation begins at the end of an ag- The long frame is passed down to the physical layer.
gregation period; The green line stands for the boundary of the EIFA
3) The generated long frame is transmitted upon scheme.

The upper layers

Traffic from the network layer

Network layer

Data buffer
Short frame Short frame Short frame Short frame
#1 #2 #3 #k
MAC layer
The EIFA scheme
Up to k short frames are aggregated
Frame at every aggregation period
aggregation with duration of T
Boundary

Physical layer Long frame

Fig. 2. Diagram of the EIFA.

Recall that IEEE 802.11ax standard, or Wi-Fi 6, 1) Assumptions: Given a STA in the WLAN, we
aims at high efficiency (HE) [7], which differs from assume that the short frame arrivals at the data buffer
IEEE 802.11ac standard (i.e., Wi-Fi 5) targeting at of the STA obey a Poisson process with rate of λ [40],
very high throughput (VHT). Only the HE STAs can [41].
take advantage of the new features introduced in Wi-Fi In order to avoid the data buffer growing without
6, including the OFDMA mechanism. Therefore, it is bound, we require the aggregation period of T to satisfy
required that the STAs operating with the EIFA be HE
k
stations so that the EIFA can achieve the best perform- T < (1)
λ
ance. In the case when each of the STAs with the EIFA
is not HE station, the EIFA cannot benefit from the The reason is explained as follows. In the EIFA
new features in Wi-Fi 6. scheme, one long frame, which consists of at most k
2. Mathematics model for the EIFA scheme short frames, is transmitted in an aggregation period of
In this section, we use Markov chain to model the T . Hence, the transmission rate of long frame, i.e., the
EIFA scheme in Fig.2. The Markov model aims at de- number of long frames being transmitted per unit of
riving the statistics (i.e., throughput and packet delay, time, is 1/T . Equivalently, the transmission rate of
etc.) under the EIFA scheme, from which we obtain the short frame is up to k/T . Therefore, we need the short
optimal pair of parameters k and T for the proposed frame transmission rate being greater than the short
EIFA scheme by solving the optimal problem that max- frame arrival rate λ. That is, k/T > λ, leading to (1).
imizes the throughput subject to the constraint on the 2) State probabilities: Let an denote the probabil-
packet delay. ity that there are exact n short frames arriving at the
752 Chinese Journal of Electronics 2023

STA during an aggregation period with duration of T . Similarly, state transitions related to S1 and Sk are
Then, we have [42] shown in Fig.3(b) and (c), respectively. Moreover, those
related to state Sj (j > k) are shown in Fig.3(d). The
(λT )n −λT
an = e , n = 0, 1, . . . (2) explanation of Fig.3(b)–(d) is similar to Fig.3(a), which
n!
is omitted.
The STA is said to be in state Sj if there are j From Fig.3, we obtain all state transition probabil-
short frames in the data buffer at the STA (j = 0, 1, ities as follows:
. . .). Then, the state space is Ω = {S0 , S1 , . . .}. Let state 1) For i = 0, 1, . . . , k ,
probability Pj = Pr{Sj }, which is the probability of the ∑
event that the STA is in state Sj ∈ Ω . Thus, it holds  k−i a , j = 0
h
∑ qi,j = h=0 (5)
that Sj ∈Ω Pj = 1. Equivalently, 
ak+j−i , j = 1, 2, . . .
∞
∑
Pj = 1 (3) 2) For i = k + 1, k + 2, . . . ,
j=0 {
0, j = 0, 1, · · · , i − k − 1
In addition, for two states Si , Sj ∈ Ω , we intro- qi,j = (6)
aj−i+k , j = i − k, i − k + 1, . . .
duce notation qi,j = Pr{Sj |Si } to represent the probab-
ility of transition from Si to Sj . Now, we start deriving the state probability Pj =
Then, we consider the state transitions occur at the Pr{Sj }, j = 0, 1, . . . . The event of the STA being in
ends of aggregation periods, with the assumption that state Sj ∈ Ω is equivalent to the union of the following
the STA is initialized to be in state S0. events:
State transitions related to S0 are illustrated in 1) The STA is originally in state Sj and then it re-
Fig.3(a), which is explained as follows. We first con- mains the same state in an aggregation period, which
sider the state transitions starting from S0. During an occurs with probability of Pj qj,j ; and
aggregation period with length of T , the number of 2) The STA is originally in state Sh ∈ Ω other
short frames arriving at the STA is possibly 0, 1, . . . , than Sj (i.e., h ̸= j) and then it changes its state from
k, . . . with probability of a0 , a1 , . . . , ak , . . . , respectively. Sh to Sj , which occurs with probability of Ph qh,j , Sh ∈ Ω−
Due to the fact that at most k short frames are aggreg- {Sj }.
ated to form one long frame being transmitted at the Thus, we have
end of the aggregation period, the event that the num-
ber of frame arrivals being 0, 1, . . . , or k , which hap- ∞
∑
∑k Pj = Ph qh,j , j = 0, 1, . . . (7)
pens with probability of i=0 ai , does not change the
h=0
STA’s state. That is, all the arrived short frames are
aggregated and transmitted so that no short frame is The above equations can be expressed as follows:
left at the data buffer. Hence, the transition from state
∑k
S0 to itself happens with probability of i=0 ai . A∞ P⃗∞ = ⃗0∞ (8)
However, k + 1, k + 2, . . . frame arrivals cause the STA’s
state to be S1 , S2 , . . . with probability ak+1 , ak+2 , . . . , where P⃗∞ = (P0 , P1 , . . . , Pn , . . .)T, ⃗0∞ is a vector with
respectively. Then, we consider state transitions that infinite elements being 0 s, and
end at S0. In the case when i = 1, 2, . . . , k , the STA’s  
q0,0 − 1 q1,0 ··· qn,0 ···
state changes from Si to S0 when no more than k − i
 q0,1 q1,1 − 1 ··· qn,1 · · ·
frames arrives at the STA, which occurs with probabil-  . .. .. .. .. 
∑k−i A∞ =
 .. . . . . (9)
ity of j=0 aj, because no frame stays at the data buf-  q0,n q1,n ··· qn,n − 1 · · ·
fer after the aggregation process that consumes at most .. .. .. .. ..
. . . . .
k short frames. Clearly, transition from states
Sk+1 , Sk+2 , . . . to S0 does not occur in one frame aggreg- Equation (3) indicates that the infinite series
∑∞
ation period under the proposed EIFA scheme. j=0 Pj converges to 1, which requires limn→∞ Pn = 0.
Thus, transition probabilities out of state S0 can be As a result, for any small real number ε , say ε = 10−4,
represented by there exists a sufficient large integer M so that
∑
 k a, j=0 PM +1 + PM +2 + · · · < ε (10)
i
q0,j = i=0 (4)

ak+j , j = 1, 2, . . . which indicates that the probability of the data buffer
 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 753

ak+k+1, ak+k+2,...

ak+k
ak+k−1

k
a
··+
ak+2

+·
1
+a
0
S0
ak+1
S1 S2 Sk−1 Sk Sk+1
a

a0+a1+···+ak−1 a0+a1+···+ak−2 a1 a0

(a) State transitions with state S0

ak+k+1, ak+k+2,...

ak+k
ak+k−1
ak+k−2
ak
ak+1 ak+1
S0 S1 S2 Sk−1 Sk Sk+1
ak−1 a2 a1 a0
a0+a1+···+ak−1

(b) State transitions with state S1

ak+2, ak+3,···
ak+k ak+k−1 ak+1

ak+1
S0 S1 Sk−1 ak Sk Sk+1

a1 ak−1 ak−1

a0

(c) State transitions with state Sk

ak+2, ak+3,···
ak+j ak+j−1
aj+1

ak+1
Sj−k Sj−k+1 Sj−1 ak Sj Sj+1

a1 ak−1 ak−1
a0

(d) State transitions with state Sj when j > k

Fig. 3. State transition diagram.

having more than M short frames can be ignored. Equi- tions in (8) can be approximated with precision of ε by
valently, the following M + 1 equations:

ÃM P⃗M = ⃗0M (12)

∑
M
Pj ≈ 1 (11)
j=0
In (12), both P⃗M and ⃗0M are (M + 1)-dimension
vectors and ÃM is an (M +1) × (M +1) matrix, which
with precision of ε . are defined as follows: P⃗M = (P0 , P1 , . . . , PM )T and ⃗0M
Therefore, with this M , the set of infinite equa- has all elements of 0 s, and
754 Chinese Journal of Electronics 2023
q − 1  
0,0 q1,0 ··· qM,0  1 1 1 1 1 ... 1 1
 q0,1 q1,1 − 1 ··· qM,1   a0 
ÃM = . .. .. .. . (13)  
.. .  a1 a0 
. .  
q0,M q1,M ··· qM,M − 1  a2 a1 a0 
 
 .. .. .. 
Usually, matrix ÃM is not fully ranked, which may  . . . 
 
 ak−1 ... a2 a1 a0 
cause the problem that the set of equations in (12) does  
 ak ak−1 ... a2 a1 a0 
not have unique solution. Fortunately, this problem can  
 .. .. .. 
be remedied by replacing one of the equations, say the  . . . 
 
first equation, with (11). Then, we have aM −k−1 aM −k−2 ... a3 a2 a1 a0 
 
 aM −k aM −k−1 ... a3 a2 a1 
 
AM P⃗M = ⃗bM (14) aM −k+1 aM −k ... a3 a2 
 
 .. .. 
 . .
where ⃗bM = (1, 0, 0, . . . , 0)T and the (M + 1) × (M + 1)  
 aM −2 aM −3 ... ak+2 ak+1 ak ak−1
matrix AM is
aM −1 aM −2 ... ak+3 ak+2 ak+1 ak
 1 1 ··· 1 
Equations (14) and (16) lead to
 q0,1 q1,1 − 1 ··· qM,1 
AM = . .. .. ..  (15) ( )−1
.. . . . P⃗M = BM 1 − BM 2 ⃗bM (19)
q0,M q1,M ··· qM,M − 1
from which we obtain the state probabilities.
Using (5) and (6) in (15), we obtain
3. Statistics of the EIFA scheme
AM = BM 1 − BM 2 (16) With the state probabilities in hands, the expected
number of short frames in the data buffer at the STA
Here, both BM 1 and BM 2 are (M + 1) × (M + 1) can be approximated by
matrices, where
∑
M
N̄ = jPj (20)
BM 2 = diag{0, 1, 1, · · · , 1}
| {z } (17) j=0
M
with precision of ε .
and BM 1 consists of two (M + 1)-row matrices BM 11 The buffer overflow probability of the EIFA
and BM 12 as scheme is
[ ]
BM 1 = BM 11 : BM 12 (18) ∑
M
p̂ = Pj (21)
j=B+1
In the above equation, BM 11 is an (M +1) × (k+1)
matrix as where B is the buffer size measured in short frames.
  4. Throughput
1 1 1 1 ... 1 Now, we begin to measure the throughput in the
 ak+1 ak ak−1 ak−2 ... a1 
  EIFA scheme. Let L be the size of a frame (in bits).
 ak+2 a a a . . . a2 
 k+1 k k−1  In the case when there are n short frames in the
 ak+3 ak+2 ak+1 ak ... a3 
  data buffer at the STA, which happens with probabil-
 .. .. .. .. .. .. 
 . . . . . .  ity of Pn, we represent these frames by ξ1 , ξ2 , . . . , ξn .
 
 a2k a2k−1 a2k−2 a2k−3 ... ak  It holds that
 
 a2k+1 a2k a2k−1 a2k−2 ... ak+1 
 
 .. .. .. .. .. ..  n = mk + r (22)
 . . . . . . 
 
 aM a a a . . . a  where m is a non-negative integer and r is such an in-
 M −1 M −2 M −3 M −k 
 aM +1 a a a . . . a  teger that 0 ≤ r < k .
 M M −1 M −2 M −k+1
 aM +2 a a a . . . a  We divide the above n frames into m + 1 groups as
 M +1 M M −1 M −k+2
 . . . . . .  {ξ1 , ξ2 , . . . , ξk }, {ξk+1 , ξk+2 , . . . , ξ2k }, {ξ2k+1 , ξ2k+2 , . . . ,
 .. .. .. .. .. .. 
 
aM +k−1 aM +k−2 aM +k−3 aM +k−4 . . . aM −1  ξ3k }, . . . , {ξ(m−1)k+1 , ξ(m−1)k+2 , . . . , ξmk }, and {ξmk+1 ,
aM +k aM +k−1 aM +k−2 aM +k−3 . . . aM ξmk+2 , . . . , ξn−1 , ξn }.
The last group is with r frames, but each of the
and BM 12 is an (M + 1) × (M − k) matrix as other groups contains k frames. Thus, each group can
 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 755

be aggregated and transmitted in a single aggregation Here, H stands for the sum of the time expended in
period. Hence, in the case when r ̸= 0, i.e., integer n is contending for the channel, the preamble, headers of
not dividable by integer k , these n frames, which carry the physical layer and the MAC layer, and the frame
nL bits, cause the STA to totally consume m + 1 ag- check sequence (FCS), together with the time for re-
gregation periods and spend duration of (m + 1)T in ceiving the ACK frame; r is the data rate applied by
data delivery. Additionally, in the case when r = 0, it the STA in delivering the aggregated frame, i.e., a long
consumes m aggregation periods. frame consisting of up to k short frames.
In summary, the STA expends ⌈ nk ⌉T aggregation In the above optimization problem, the constraint
periods in delivering the n frames, where ⌈·⌉ is the ceil (25b) combines (1) with the equation H + kL/r < T ,
( )
function. As a result, nL/ ⌈ nk ⌉T bits are delivered per required for a long frame to be transmitted in one ag-
unit of time. gregation period. In constraint (25c), d is a preset con-
Therefore, the expected throughput can be approx- stant as the upper bound of frame delay. And the con-
imated with precision of ε by straint (25d) reflects how many short frames can be ag-
gregated to generate a long frame, which is regulated in
∑
M
nL L∑
M
1 IEEE 802.11ax standard [7].
θ(k, T ) = Pn n = nPn n (23) Next, we consider solving the optimization prob-
n=1 ⌈ ⌉T T n=1 ⌈ ⌉
k k lem.
Let σ be a unit of time, say a millisecond (ms).
In the sequel, we refer to θ(k, T ) as the approxim-
Then, let
ated throughput.
⌈ ( )⌉
5. Packet delay 1 kL
t1 = H+ (26)
Next, we study frame delay in the EIFA scheme. σ r
Delay of a frame is defined as the time interval from
the instant when it arrives at the data buffer at the and
STA to the instant when it has been aggregated and ⌊ ⌋
k
transmitted. A frame may arrive at the buffer at any t2 = (27)
λσ
time in an aggregation period. Hence, the interval from
the arrival time to the end of the aggregation period Here, t1 and t2 take the multiples of the time unit
with duration of T takes a random value in [0, T ]. We σ , and act as the lower bound and the upper bound for
approximate it with the average of T /2. Thus, in the the variable T in Constraint (25b).
case when the frame arrives at the buffer having n buf- Then, the optimization problem is transformed in-
fered frames, its delay can be expressed by T2 + ⌈ n+1 to the following one with the variables k and T being
k ⌉T ,
where the second item stands for the time spent by the positive integers.
STA on aggregating and transmitting the previous n max θ(k, T ) w.r.t. k, T (28a)
frames in the buffer, together with the newly arrived
frame. Hence, the expected delay of the frame is ap- s.t. k ∈ {2, 3, . . . , 64}, (28b)
proximated with precision of ε by T ∈ T = {t1 , t1 + 1, t1 + 2, . . . , t2 }, (28c)
(1 (28d)
n+1 )
∑
M τ (k, T ) < d
τ (k, T ) = T Pn +⌈ ⌉ (24)
n=0
2 k In the case when there are a few elements in the
set T in (28c), the optimization problem (28a) can be
IV. Optimization Problem solved by the exhaustive search. Otherwise, we can ap-
ply genetic algorithm that consists of selection, crossov-
To help our EIFA achieve the maximum through- er, mutation operations [43], which is omitted due to
put, we formulate the optimization problem in (25a) space limitation.
that maximizes the approximated throughput under the
constraint on the approximated frame delay. V. Performance Evaluation
max θ(k, T ) w.r.t. k, T (25a) In this section, we evaluate performance of the pro-
kL k posed EIFA via simulation that mimics the OFDMA
s.t. H + <T < , (25b) with UORA. The parameters applied in the simulation
r λ
are shown in Table 1, where Ttr is the duration of a
τ (k, T ) < d, (25c)
trigger frame, TSIFS stands for short inter-frame space
k ∈ {2, 3, . . . , 64} (25d) (SIFS), Tpre is the duration of preamble in the physical
756 Chinese Journal of Electronics 2023

layer, TPHYhdr is the duration of header of the physical 80 d=0.01 (s)

EIFA
Maxk
layer, and TBA represents the duration of block ACK. In 70 Mink
Randk
the simulation, we mimic the duration of 107 short 60

Throughput (Mbps)
Twolevel
frame arriving to a STA. 50
40
Table 1. Parameters and values
30
Parameter Value
r 20
200 Mbps
L 1500 Bytes 10
Ttr 100 μs 0
TSIFS 16 μs 1000 3000 5000 7000 9000
λ (packets/s)
Tpre 36 μs
(a) Throughput
TPHYhdr 40 μs
TBA 40 μs 0.08 d=0.01 (s) EIFA
ε 0.01 Maxk
0.07 Mink
Randk
0.06 Twolevel
The proposed EIFA is compared with the frame 0.05

Delay (s)
aggregation schemes “Maxk”, “Mink”, and “Randk”, 0.04
where “Maxk” and “Mink” always let the STA choose
0.03
the maximum NoS, i.e., 64, and the minimum NoS, i.e.,
0.02
2, respectively, and “Randk” lets the STA randomly
0.01
picks an NoS in the set {2, 3, 4, . . . , 64}. Moreover, we
also compare the EIFA with the two-level frame aggreg- 0
1000 3000 5000 7000 9000
ation scheme, simply referred to as “Twolevel” scheme, λ (packets/s)
presented in [34], which maximizes throughput. We set (b) Delay
the bit error rate (BER) to 0 in the Twolevel scheme Fig. 4. Impact of packet arrival rate on throughput and
due to no BER considered in the EIFA. It should be delay (d = 0.01 (s)).
pointed out that the Twolevel scheme assumes satur-
120 d=0.02 (s)
ated traffic [34], which may not hold in practice. EIFA
Maxk
Setting delay bound d = 0.01, 0.02, 0.03 (s) and 100 Mink
Randk
short frame arrival rate λ = 1000, 3000, 5000, 7000, 9000
Throughput (Mbps)

Twolevel
80
(in packets per second), we obtain the results shown in
Figs.4, 5 and 6, where the Twolevel scheme does not 60
consider delay bound d , i.e., every packet contributes to
40
the throughput in the Twolevel scheme regardless of
whether its delay is greater than d or not. Hence, the 20
throughput surely takes a lower value if delay bound d
0
is applied. This is because the packets with the delay 1000 3000 5000 7000 9000
λ (packets/s)
longer than d are discarded.
(a) Throughput
It can be observed from Figs.4(a), 5(a) and 6(a)
that EIFA outperforms the other schemes in through- 0.08 d=0.02 (s)
EIFA
Maxk
put, which realizes the goal of the optimization prob- 0.07 Mink
Randk
lem in (25a). From these figures, we have the observa- 0.06 Twolevel
tion that the throughput in the Twolevel scheme im- 0.05
Delay (s)

proves when λ grows. In other words, heavier traffic 0.04

brings in a higher throughput in the Twolevel scheme, 0.03
which agrees with the saturated traffic assumption in
0.02
the Twolevel.
0.01
Figs.4(b), 5(b), and 6(b) reflect the fact that the
0
EIFA increases throughput at cost of packet delay. 1000 3000 5000 7000 9000
That is, the EIFA can tune the pair of parameters k λ (packets/s)
(b) Delay
and T such that throughput is improved while the pre-
set delay bound is satisfied. It can be clearly seen from Fig. 5. Impact of packet arrival rate on throughput and
these figures that, for a given λ, the Twolevel scheme delay (d = 0.02 (s)).
 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 757

120 d=0.03 (s) age and improves throughput.

EIFA
Maxk
100 Mink In summary, the proposed EIFA can improve
Randk
throughput.
Throughput (Mbps)
Twolevel
80

60
VI. Conclusions
40
Frame aggregation can improve throughput in the
WLANs based on the family of IEEE 802.11 standard.
20 The proposed EIFA scheme takes the new feature, i.e.,
0 the OFDMA mechanism, of the newest IEEE 802.11ax
1000 3000 5000 7000 9000 standard and enables the STAs in the 802.11ax WLANs
λ (packets/s)
(a) Throughput to set the optimal NoS and waiting time in frame ag-
gregation, which improves throughput. In this paper,
0.08 d=0.03 (s) EIFA
Maxk we develop a novel Markov chain to model the one-level
0.07 Mink
Randk frame aggregation. In the future, we will investigate the
0.06 Twolevel
Markov model for the two-level frame aggregation.
0.05
Delay (s)

0.04
References
0.03
[1] IEEE Std 802.11-1997:1997, Information Technology-Tele-
0.02
communications and Information Exchange Between Sys-
0.01 tems-Local and Metropolitan Area Networks-Specific Re-
0 quirements Part 11: Wireless LAN Medium Access Control
1000 3000 5000 7000 9000
λ (packets/s) (MAC) and Physical Layer (PHY) Specifications, doi:
(b) Delay 10.1109/IEEESTD.1997.85951.
[2] IEEE Std 802.15.4-2006:2006, IEEE Standard for Informa-
Fig. 6. Impact of packet arrival rate on throughput and tion Technology-Local and Metropolitan Area Networks-
delay (d = 0.03 (s)).
Specific Requirements-Part 15.4: Wireless Medium Access
has delay larger than the other schemes, even larger Control (MAC) and Physical Layer (PHY) Specifications
for Low Rate Wireless Personal Area Networks (WPANs),
than delay bound d in some cases. Therefore, the
doi: 10.1109/IEEESTD.2006.232110.
throughput in the Twolevel scheme will be much lower
[3] S. Tozlu, M. Senel, W. Mao, et al., “Wi-Fi enabled sensors
than the EIFA when delay bound d is applied in the for internet of things: A practical approach,” IEEE Commu-
Twolevel. nications Magazine, vol.50, no.6, pp.134–143, 2012.
Impact of delay bound d on throughput in the [4] Q. Q. Yang, G. Y. Zhou, W. H. Qin, et al., “Air-kare: A
EIFA is shown in Fig.7, from which we observe that, Wi-Fi based, multi-sensor, real-time indoor air quality mon-
for a given λ, relaxing delay bound helps in improving itor,” in Proceedings of 2015 IEEE International Wireless
throughput in the EIFA scheme. This agrees with our Symposium (IWS 2015), Shenzhen, China, pp.1–4, 2015.
[5] R. R. Rajanna, S. Natarajan, and P. R. Vittal, “An IoT Wi-
intuition that a higher delay bound enables the STA to
Fi connected sensor for real time heart rate variability mon-
have a longer waiting time for more short frames to
itoring,” in Proceedings of the 3rd International Confer-
come and thus more subframes can be aggregated into a ence on Circuits, Control, Communication and Computing
longer frame, which increases efficiency of channel us- (I4C), Bangalore, India, pp.1–4, 2018.
110 [6] M. Sivachitra, J. J. D. Kumar, R. Gokul, et al., “Wi-Fi con-
d=0.01 (s)
100 trolled multi-sensor robotic car,” in Proceedings of the 2nd
d=0.02 (s)
90 d=0.03 (s) International Conference on Artificial Intelligence and
Smart Energy (ICAIS), Coimbatore, India, pp.1541–1546,
Throughput (Mbps)

80
2022.
70
[7] IEEE Std 802.11ax-2021:2021, IEEE Standard for Informa-
60
tion Technology-telecommunications and Information Ex-
50
change Between Systems Local and Metropolitan Area Net-
40
works-specific Requirements Part 11: Wireless LAN Medi-
30 um Access Control (MAC) and Physical Layer (PHY) Spe-
20 cifications Amendment 1: Enhancements for high-efficiency
10 WLAN, doi: 10.1109/IEEESTD.2021.9442429.
1000 3000 5000 7000 9000
λ (packets/s) [8] D. J. Deng, Y. P. Lin, X. Yang, et al., “IEEE 802.11ax:
Highly efficient WLANs for intelligent information infra-
Fig. 7. Throughput vs. delay upper bound. structure,” IEEE Communications Magazine, vol.55, no.12,
758 Chinese Journal of Electronics 2023

pp.52–59, 2017. 2016.

[9] IEEE, IEEE Std 802.11nTM 2009:2009, IEEE Standard for [22] H. Assasa, A. Loch, and J. Widmer, “Packet mass transit:
Information Technology-Telecommunications and Informa- Improving frame aggregation in 60 GHz networks,” in Pro-
tion Exchange Between Systems-Local and Metropolitan ceedings of the IEEE 17th International Symposium on A
Area Networks-Specific Requirements-Part 11: Wireless World of Wireless, Mobile and Multimedia Networks
LAN Medium Access Control (MAC) and Physical Layer (WoWMoM), Coimbra, Portugal, pp.1–7, 2016.
(PHY) Specifications, Amendment 5: Enhancements for [23] Y. Nomura, K. Mori, and H. Kobayashi, “Efficient frame
Higher Throughput, New York, NY, USA: IEEE. aggregation with frame size adaptation for next generation
[10] D. Kotagiri, K. Nihei, and T. S. Li, “Distributed convolu- MU-MIMO WLANs,” in Proceedings of the 9th Interna-
tional deep reinforcement learning based OFDMA MAC for tional Conference on Next Generation Mobile Applications,
802.11ax,” in Proceedings of ICC 2021-IEEE International Services and Technologies, Cambridge, UK, pp.288–293,
Conference on Communications, Montreal, QC, Canada, 2015.
pp.1–6, 2021. [24] E. Charfi, L. Chaari, and L. Kamoun, “Joint urgency delay
[11] D. H. Xie, J. W. Zhang, A. M. Tang, et al., “Multi-dimen- scheduler and adaptive aggregation technique in IEEE
sional busy-tone arbitration for OFDMA random access in 802.11n networks,” in Proceedings of the 5th International
IEEE 802.11ax,” IEEE Transactions on Wireless Commu- Conference on Communications and Networking (COM-
nications, vol.19, no.6, pp.4080–4094, 2020. NET), Tunis, Tunisia, pp.1–6, 2015.
[12] Y. Zheng, J. B. Wang, Q. H. Chen, et al., “Retransmission [25] X. L. Zhou and A. Boukerche, “AFLAS: An adaptive frame
number aware channel access scheme for IEEE 802.11ax length aggregation scheme for vehicular networks,” IEEE
based WLAN,” Chinese Journal of Electronics, vol.29, no.2, Transactions on Vehicular Technology, vol.66, no.1,
pp.351–360, 2020. pp.855–867, 2017.
[13] W. J. Lee, W. Shin, J. A. Ruiz-de-Azua, et al., “NOMA- [26] M. Yazid and A. Ksentini, “Modeling and performance ana-
based uplink OFDMA collision reduction in 802.11ax net- lysis of the main MAC and PHY features of the 802.11ac
works,” in Proceedings of 2021 International Conference on standard: A-MPDU aggregation vs spatial multiplexing,”
Information and Communication Technology Convergence IEEE Transactions on Vehicular Technology, vol.67, no.11,
(ICTC), Jeju Island, Korea, Republic of, pp.212–214, 2021. pp.10243–10257, 2018.
[14] L. Lanante, C. Ghosh, and S. Roy, “Hybrid OFDMA ran- [27] J. Saldana, J. Ruiz-Mas, and J. Almodóvar, “Frame aggreg-
dom access with resource unit sensing for next-gen 802.11ax ation in central controlled 802.11 WLANs: The latency
WLANs,” IEEE Transactions on Mobile Computing, vol.20, versus throughput tradeoff,” IEEE Communications Let-
no.12, pp.3338–3350, 2021. ters, vol.21, no.11, pp.2500–2503, 2017.
[15] R. G. Cheng, C. M. Yang, B. S. Firmansyah, et al., “Up- [28] J. Saldana, O. Topal, J. Ruiz-Mas, et al., “Finding the
link OFDMA-based random access mechanism with bursty sweet spot for frame aggregation in 802.11 WLANs,” IEEE
arrivals for IEEE 802.11ax systems,” IEEE Networking Let- Communications Letters, vol.25, no.4, pp.1368–1372, 2021.
ters, vol.4, no.1, pp.34–38, 2022. [29] A. Abedi, T. Brecht, and O. Abari, “Demystifying frame ag-
[16] W. R. Ghanem, V. Jamali, Y. Sun, et al., “Resource alloca- gregation in 802.11 networks: Understanding and approxim-
tion for multi-user downlink MISO OFDMA-URLLC sys- ating optimality,” Computer Communications, vol.180,
tems,” IEEE Transactions on Communications, vol.68, pp.259–270, 2021.
no.11, pp.7184–7200, 2020. [30] M. Morimoto, K. Sanada, H. Hatano, et al., “Throughput
[17] F. Khoramnejad, M. Rasti, H. Pedram, et al., “Load man- and delay analysis for full-duplex WLANs with frame ag-
agement, power and admission control in downlink cellular gregation under non-saturated condition,” in Proceedings of
OFDMA networks,” IEEE Transactions on Mobile Com- the 22nd International Symposium on Wireless Personal
puting, vol.20, no.4, pp.1477–1493, 2021. Multimedia Communications (WPMC), Lisbon, Portugal,
[18] M. J. Xiao, L. S. Huang, K. Xing, et al., “Opportunistic pp.1–5, 2019.
data aggregation in low-duty-cycle wireless sensor networks [31] K. Liu and C. L. Li, “An efficient adaptive frame aggrega-
with unreliable links,” Chinese Journal of Electronics, tion scheme in vehicular Ad Hoc networks,” in Proceedings
vol.22, no.3, pp.599–603, 2013. of the 9th International Conference on Wireless Commu-
[19] Y. X. Lin and V. W. S. Wong, “WSN01-1: Frame aggrega- nications and Signal Processing (WCSP), Nanjing, China,
tion and optimal frame size adaptation for IEEE 802.11n pp.1–6, 2017.
WLANs,” in Proceedings of IEEE Globecom 2006, San [32] S. Kim and J. H. Yun, “Efficient frame construction for
Francisco, CA, USA, pp.1–6, 2006. multi-user transmission in IEEE 802.11 WLANs,” IEEE
[20] M. M. S. Kowsar and S. Biswas, “Performance improve- Transactions on Vehicular Technology, vol.68, no.6,
ment of IEEE 802.11n WLANs via frame aggregation in NS- pp.5859–5870, 2019.
3,” in Proceedings of 2017 International Conference on [33] J. L. Liu, M. W. Yao, and Z. L. Qiu, “Enhanced two-level
Electrical, Computer and Communication Engineering frame aggregation with optimized aggregation level for
(ECCE), Cox’s Bazar, Bangladesh, pp.1–6, 2017. IEEE 802.11n WLANs,” IEEE Communications Letters,
[21] S. Abdallah and S. D. Blostein, “Joint rate adaptation, vol.19, no.12, pp.2254–2257, 2015.
frame aggregation and MIMO mode selection for IEEE [34] K. Suzuki and S. Yamazaki, “Throughput maximization
802.11ac,” in Proceedings of 2016 IEEE Wireless Commu- based on optimized frame-aggregation levels for IEEE
nications and Networking Conference, Doha, Qatar, pp.1–6, 802.11 WLANs,” IEEE Communications Letters, vol.25,
 Enhancing Network Throughput via the Equal Interval Frame Aggregation Scheme for IEEE 802.11ax WLANs 759

no.5, pp.1725–1728, 2021. Academic Press, New York, NY, USA, doi: 10.1016/C2012-
[35] B. S. Kim, H. Y. Hwang, and D. K. Sung, “Effect of frame 0-03564-8, 2014.
aggregation on the throughput performance of IEEE [43] J. H. Holland, Adaptation in Natural and Artificial Sys-
802.11n,” in Proceedings of 2008 IEEE Wireless Commu- tems, The MIT Press, Cambridge, MA, USA, 1992.
nications and Networking Conference, Las Vegas, NV,
USA, pp.1740–1744, 2008. ZHU Yihua received the B.S.
[36] S. Seytnazarov, J. G. Choi, and Y. T. Kim, “Enhanced degree in mathematics from Zhejiang
mathematical modeling of aggregation-enabled WLANs with Normal University, Zhejiang, China, in
compressed BlockACK,” IEEE Transactions on Mobile July 1982; the M.S. degree in operation
Computing, vol.18, no.6, pp.1260–1273, 2019. research and cybernetics from Shanghai
[37] I. Jabri, K. Mansour, I. Al-Oqily, et al., “Enhanced charac- University, Shanghai, China, in April 1993;
terization and modeling of A-MPDU aggregation for IEEE and the Ph.D. degree in computer sci-
ence and technology from Zhejiang Uni-
802.11n WLANs,” Transactions on Emerging Telecommu-
versity, Zhejiang, China, in March 2003.
nications Technologies, vol.33, no.1, article no.e4384, 2022.
Dr. Zhu is a Professor at Zhejiang University of Technology,
[38] W. W. Lu, S. L. Gong, and Y. H. Zhu, “Timely data deliv-
Hangzhou, Zhejiang, China. He is a Member of China Computer
ery for energy-harvesting IoT devices,” Chinese Journal of
Federation Technical Committee on Internet of Things (IoT). His
Electronics, vol.31, no.2, pp.322–336, 2022.
[39] D. Skordoulis, Q. Ni, H. H. Chen, et al., “IEEE 802.11n current research interests include IoT, WLANs, WSNs, RFID sys-
MAC frame aggregation mechanisms for next-generation tems. He has served as Technical Program Committee Members
high-throughput WLANs,” IEEE Wireless Communica- or Co-chairs in the international conferences IEEE ICC, WCNC,
tions, vol.15, no.1, pp.40–47, 2008. GlobeCom, DCOSS, etc. (Email: [email protected])
[40] Y. H. Zhu and V. C. M. Leung, “Efficient power manage-
XU Mengying received the B.S.
ment for infrastructure IEEE 802.11 WLANs,” IEEE
degree from Zhejiang A & F University,
Transactions on Wireless Communications, vol.9, no.7,
Zhejiang, China, in 2020. She is cur-
pp.2196–2205, 2010.
rently a candidate student for the M.S.
[41] Q. H. Chen and Y. H. Zhu, “Scheduling channel access degree in Zhejiang University of Techno-
based on target wake time mechanism in 802.11ax logy. Her current research interests in-
WLANs,” IEEE Transactions on Wireless Communica- clude channel access in the IEEE 802.11ax
tions, vol.20, no.3, pp.1529–1543, 2021. WLAN.
[42] S. M. Ross, Introduction to Probability Models, 11th ed.,