Professional Documents
Culture Documents
Full Duplex Wireless Communiation With Dynamic Buffer Aided MAC Protocol
Full Duplex Wireless Communiation With Dynamic Buffer Aided MAC Protocol
Regular paper
A R T I C L E I N F O A B S T R A C T
Keywords: This paper presents a dynamic buffer aided medium access control protocol for asymmetric traffic conditions in
Full duplex an IoT network, where the IoT devices send data to a wireless access point (WAP) which then forwards the data
Cross layer to the network fusion center. Specifically, the control frame structures of conventional carrier sense multiple
MAC protocol
access with collision detection (CSMA/CD) protocol are modified for incorporating the traffic asymmetry of the
network. The devices and WAP are assumed to have full duplex abilities to support continuous sensing of the
channel. The sensing performance of the considered system is derived for arbitrary number of devices in terms of
probabilities of false alarm and miss detection. Moreover, the throughput of the network is also obtained.
Simulation results show that the proposed cross-layer design achieves significant throughput improvement while
incorporating traffic asymmetry of the network.
* Corresponding author.
E-mail address: anshul.maloo23@gmail.com (A. Maloo).
https://doi.org/10.1016/j.aeue.2020.153488
Received 19 March 2020; Accepted 1 October 2020
Available online 19 October 2020
1434-8411/© 2020 Elsevier GmbH. All rights reserved.
A. Maloo and P.K. Sharma AEUE - International Journal of Electronics and Communications 128 (2021) 153488
• Proposed protocol considers the traffic asymmetry between the WAP for the ith user it is denoted as ri .
to NFC and incorporates the dynamic buffer in WAP which may
change dynamically depending on the variable traffic conditions. 3. Dynamic buffer aided protocol
• In order to increase the throughput and to avoid the hidden terminal
problem, a new framing structure is proposed. The main idea behind this protocol is to incorporate the asymmetric
• A practical scenario is considered where an arbitrary number of N traffic conditions which may arise in case many WAPs communicate
devices are present. In this scenario sensing performance is analysed with NFC at different rates simultaneously. In the following we discuss
with Rayleigh faded RSI and the expression for probability of false the frame design and proposed protocol (see Fig.1).
alarm (pfa ), probability of miss detection (pmd ) and throughput are
derived. 3.1. Frame Design
The rest of the paper is organized as follows. In Section 2, the In this protocol we use five different types of control frames, U-RTS
physical layer system model is given. Section 3 describes the proposed (user-request to send), FD-CTS (FD-clear to send), U-CTS, FD-RTS, and
protocol. The cross-layer performance is analysed in Section 4. Section 5 ACK (acknowledgement) as shown in Fig.2. Moreover, a modified
presents numerical results and the paper is concluded in Section 6. The header is also considered in the proposed protocol. The description of
notations used in this paper are summarized in Table 1. each of the control frames is given below:
2. Physical layer system model 1. U-RTS and FD-RTS: The U-RTS and FD-RTS frames are used by user
and WAP respectively. In U-RTS frame, an additional field of buffer
A scattered network, consisting of an arbitrary number N of size is added. The additional field includes buffer state information of
randomly distributed IoT devices which communicate to the NFC the user which indicates the number of packets to be transmitted by
through a WAP, is considered. The ith and jth IoT devices are represented the user. This information helps WAP acknowledge the last packet in
by Ui and Uj , respectively, where i, j ∈ {1,2,⋯,N}, and are assumed to lie the burst transmission rather than acknowledging every packet in
in the WAP coverage area of radius R. The distance between Ui and Uj is transmission.
di,j , and di,j << di,AP , where di,AP represents the distance between ith 2. FD-CTS and U-CTS: The FD-CTS and U-CTS frames are used by WAP
device and WAP. In order to reduce the delay and to enable the energy and user respectively. FD-CTS frames includes information of
detection based continuous sensing, the IoT nodes are assumed to have contention window (CW) along with buffer size of WAP. CW aligns
FD abilities. We further assume that the traffic demand in the network is the time period for data transmission between WAP and user. Also in
surplus and all the devices have their data for transmission, however the FD-CTS, information of dynamic buffer size at WAP is received by
WAP can help only one of the users transmit its data to NFC. The channel user which helps incorporate the effect of traffic conditions between
access among the multiple users is assumed to be performed on the basis WAP and NFC.
of proposed dynamic buffer aided MAC protocol. At physical layer, the 3. Header of user: It contains two fields related to device i.e. its
channel among ith user and WAP is represented by hi,AP and is assumed to maximum buffer size and its rank. Rank is used by WAP to refer to a
be distributed as the Rayleigh faded random variable. The RSI at the FD device which can give the information of overheads using maximum
users is modeled as the Rayleigh distributed random variable [9,10] and buffer size of the user.
2
A. Maloo and P.K. Sharma AEUE - International Journal of Electronics and Communications 128 (2021) 153488
1. Control frames transmission: This stage consists of two processes, is generated by WAP and is given as BAP = Lf + Lt , where Lf is the
first, the sensing and contention and second, the frame free buffer size of WAP and Lt is length of data which is transmitted
communication. to NFC in current time slot which can be given as ℛ × τ. Thus user
• Sensing and contention: When a user has the data for transmission, gets the information of available buffer at WAP through BAP .
it first senses the channel for a period of τ duration. If the channel is Step 3: After DIFS, the user transmits data according to BAP
found idle, the user waits for distributed inter frame space (DIFS) received. The user adjusts the transmission rate as per buffer in
interval, and generates a random back off time for additional delay formation received from WAP in order to synchronize the trans
before transmission. This back off time is expressed as (R(CW) × mission. The same procedure is followed to set up secondary
τ), where, R(CW) is random number of time slots over {0, 1, ⋯, transmission when WAP first gets access to channel for FD
CW} and CW is contention window which is set as 2C − 1 where C is communication.
back off stage denoted by any positive integer, C ∈ {0, 1, ⋯, Cmax }. 2. Data Transmission: The data transmission from user is completed
The user suspends the back off timer prematurely if the channel is when the user receives the acknowledgement (ACK) from WAP. The
sensed busy. Otherwise, at the end of back off timer the user gets exchange of ACK frames requires network allocation vector (NAV).
the access for the channel and starts the control frame trans When control frames are exchanged then both user and WAP get to
mission. If the packet is detected from another user at the time of know the duration field of the data frames and set the NAV accord
transmission, user stops the transmission and immediately goes to ingly. So at the end of NAV, ACK frames are sent. If ACK is not
back off procedure and back off stage is set to min(C + 1, Cmax ). received successfully, user transmits the data again.
• Frame communication process: The procedure for the transmission
is as follows: 4. Performance analysis
Step 1: At first time instant user sends U-RTS to the WAP and at
the same time instant previously stored data in the buffer of WAP, 4.1. Performance Analysis of Carrier Sensing
if any, is sent to NFC by WAP with transmission rate of ℛ.
Step 2: At second time instant, WAP sends FD-CTS which con Received signal during sensing phase can be given as,
tains its CW and variable buffer size (BAP ). BAP varies dynamically
according to the uplink transmission rate of WAP. The value of BAP
3
A. Maloo and P.K. Sharma AEUE - International Journal of Electronics and Communications 128 (2021) 153488
⎧ √̅̅̅̅̅ ∫b
1
⎨ Pi ri xi + ni ℋ0 where Γ(a, b) = Γ(a) 0 t a− 1 e− t dt is incomplete gamma function [11]. On
Y = √̅̅̅̅̅ ∑N− 1 √̅̅̅̅̅ (1)
⎩ Pi ri xi + Pj hi,j xj + ni ℋ1 . simplifying Eq. (9) and deriving the value of γ and substituting it in Eq.
j∕
=i
(10), we get
⎛ ⎞
For Ui user, ri , ni and xi are the SI channel coefficient, additive white
Gaussian noise and transmitted symbol, respectively. hi,j denotes Ray ⎜ [Γ− 1 (nt , 1 − pfa )][σ2ri + σ2μ ] ⎟
pmd = Γ⎜
⎝nt , ( ( ) )⎟ (11)
leigh fading channel coefficient between transmitting user Ui and user Uj ∑N− 1 d α ⎠
σ2ri + σ 2μ + Pref
where i, j ∈ {1, 2, ⋯, N}. Also Pi and Pj denotes transmitted power of i∕
=j di,j
P ⋅d−
where x = Γ− 1 (nt , 1 − pfa ) σref2 +σ2 and
α
where, σ2μ denotes noise power and σ2ri = χ 2 Pi [6] and χ depends upon SI μ ri
Since the channels are Rayleigh fading channels, so sum of their sam In this section we analyse the throughput of the proposed protocol.
pling power (K) is gamma distributed. Probability density function We compute the transmission probability (pt ) of every device in any time
(PDF) of K can be given as slot along with probability of successful transmission. This analysis is
1 z(nt − 1) e ϕ
− z
based on Bianchis model[12] where pt is given as
fK (z) = ⋅ (7)
ϕnt Γ(nt ) 2(1 − 2pc )
pt = (16)
(1 − 2pc )(CW + 1) + pc ⋅CW(1 − (2pc )C )
where the parameter ϕ is given as
⎧ σ2 + σ2 ℋ0 where pc is the probability that packet collides in given time slot. So
⎪ μ
⎨ ri probability of devices finishing transmission without awareness of
ϕ= ∑N− 1 ( )α (8) collision in random time slot is given as pnc = 1 − pc . Let us consider the
⎪
⎩ σ 2 + σ2 + Pref d
ri μ ℋ1 . length of data packet as L. By considering Markov chain for discrete time
i∕
=j di,j
for the back off process and sensing errors which are possible during
We use a typical path loss Rayleigh fading channel as in [8], with transmission pnc can be derived as [8]
( )α
average power as Pref ddi,j where α denotes the path loss exponent and pnc = (1 − pt )N− 1 (1 − pfa )L
Pref is the reference received signal power at reference distance d. The pfa (1 − pfa )L − p2md (17)
+(N − 1)pt (1 − pt )N− 2 pmd .
and pmd can be written as, 1 − pfa − p2md
( )
pfa = 1 − Γ nt , 2
γ
(9) Since the back off mechanism in our protocol is same as that of con
σri + σ 2μ ventional 802.11 WLAN therefore Markov model remains same for our
⎛ ⎞ protocol. Therefore successful transmission probability can be given as
⎜ γ ⎟ ps = Npt (1 − pt )N− 1
(18)
pmd = Γ⎜
⎝nt , ( ( )α ) ⎟
⎠ (10)
∑N− 1
and probability that the channel is empty is given as
d
σ + σμ + Pref
2
ri
2
i∕
=j di,j
pe = (1 − pt )N (19)
4
A. Maloo and P.K. Sharma AEUE - International Journal of Electronics and Communications 128 (2021) 153488
Therefore net collision probability can be written as proposed protocol outperforms the previous related works in terms of
network throughput. This is due to our optimized framing structure
pnet
c = 1 − Npt (1 − pt )
N− 1
− (1 − pt )N . (20) which incorporates the dynamic buffer size. Moreover, FD-WiFi lags in
throughput due to the use of conventional CSMA/CD frame structure.
We can obtain the pt by substituting the value of pfa and pmd in (17). The For the fairness purpose we only included synchronised protocols in
throughput for a device when the channel is occupied for successful throughput comparisons.
transmission can be given as In Fig. 4b shows the plot between normalized throughput with
respect to transmission rate and maximum backoff stage when pfa is
E[Payload Length of device]
⊤i = 0.01. It can be seen that the throughput performance for proposed
E[Transmission Time for payload]
(21) protocol is higher than FD-WiFi [8] and collision does not happen at
ps ⋅LUi lower back off stage. However throughput decreases at higher back off
=
pe + ps (LUi + DIFS) + pnet
c (Lc + DIFS) stage as the length of successful transmission is 50 and after that false
alarm happens due to which users go to higher contention window.
where LUi is mean length of successful transmission and given as [12] Therefore, unsuccessful transmission at higher back off stage occurs and
throughput falls rapidly.
1 − (1 − pfa )L− 1
L Ui = + (1 − pfa )L− 1
(22) In Fig. 5a we show normalized throughput versus probability of false
pfa
alarm with different network transmission data rates when N is 25 and
Cmin and Cmax are set at 3 and 8, respectively. It can be seen that the
and Lc is average length of collision and given as
throughput increases at lower pfa and attains a maximum value and then
( ) ( 1) )
N p2 1 − p2(L− decreases further at higher pfa . Result show that there exist an optimal
Lc = 1 + p2t (1 − pt )N− 2 mdnet md
. (23)
N− 2 pc (1 − p2md ) values of pfa for the proposed protocol to achieve maximum throughput.
If we increase the data transmission rate our optimal values of
We can now analyse the throughput of the proposed protocol using LUi , throughput increases. This is due to the fact that length of successful
Lc and substituting them in (21). It can be seen in the analysis that the transmission depends upon pfa and data length L.
collision and successful transmission lengths are the key factors for In Fig. 5b we can see that the normalized throughput increases
improving throughput. monotonously as compared to FD-WiFi[8] and FD-MAC. It is due to the
fact that FD-MAC [5] is not synchronized thus it leads to higher collision
5. Numerical results of data as we increase data rate. The performance of proposed protocol is
also better due to the fact that as we increase transmission speed, syn
5.1. Simulation Conditions chronization of data frames becomes important and the dynamic buffer
size and role of contention window in framing structure becomes more
Nodes in the network topology are randomly distributed in radius (R) critical. Due to this, the proposed protocol performs better in increasing
of 15 m. Transmitting power of node is set to 24 dBm. Packet size is set to the transmission rate of WAP.
1500 bytes and may vary according to simulation of different parame In Fig. 6 we plot the average end to end delay against the number of
ters. DIFS is set to standard 2 slots (30 μs). Frame slots are fixed at nodes in the network. It can be seen that the proposed protocol out
standard 40 μs. Furthermore, nt is set as 100. CWmin and CWmax are set at performs other previous related works by achieving up to 35% lower
1 and 31, respectively. delays. This is due to proposed optimized framing structure and use of
burst transmission mode. For the fairness purpose, it can be observed
5.2. Simulation Results that results in Fig. 6 match to the scenario when the packet does not
suffer any failure in the path.
In Fig. 4a the network throughput [1,8] is plotted for different
number of users considering different path loss values. Path loss expo
nent ’α’ values are assumed in range 2 to 3.5 [13]. We can see that
Fig. 4. Performance evaluation of Protocols: Throughput vs [(a) number of nodes; (b) back off stage].
5
A. Maloo and P.K. Sharma AEUE - International Journal of Electronics and Communications 128 (2021) 153488
Fig. 5. Performance evaluation of Protocols: Throughput vs [(a) false alarm probability; (b) rate of AP].
References
[1] Hu J, Di B, Liao Y, Bian K, Song L. Hybrid mac protocol design and optimization for
full duplex wi-fi networks. IEEE Trans Wireless Commun 2018;17(6):3615–30.
https://doi.org/10.1109/TWC.2018.2810119.
[2] Kim D, Lee H, Hong D. A survey of in-band full-duplex transmission: From the
perspective of phy and mac layers. IEEE Commun Surveys Tuts 2015;17(4):
2017–46. https://doi.org/10.1109/COMST.2015.2403614.
[3] Choi W, Park J, Kim Y, Sabharwal A, Lim H. Design and implementation of a full-
duplex pipelined mac protocol for multihop wireless networks. IEEE Access 2017;5:
14930–42. https://doi.org/10.1109/ACCESS.2017.2731368.
[4] Sahai A, Patel G, Sabharwal A. Pushing the limits of full-duplex: Design and real-
Fig. 6. Performance evaluation of Protocol: Average end to end delay vs time implementation. CoRR abs/1107.0607. arXiv:1107.0607.
number of nodes. [5] Tamaki K, ARH, Sugiyama Y, Bandai M, Saruwatari S, Watanabe T. Full duplex
media access control for wireless multi-hop networks. In: Proc. IEEE 77th Veh.
Technol. Conf. (VTC Spring); 2013. p. 1–5. doi:10.1109/VTCSpring.2013.6692573.
6. Conclusions [6] Al-Kadri MO, Aijaz A, Nallanathan A. X-fdr: A cross-layer routing protocol for
multihop full-duplex wireless networks. IEEE Wireless Commun 2019;26(2):70–7.
https://doi.org/10.1109/MWC.2017.1700243.
In this paper we proposed a MAC protocol with a new frame structure [7] Liao Y, Bian K, Song L, Han Z. Full-duplex mac protocol design and analysis. IEEE
that incorporates traffic asymmetry of network with the help of dynamic Commun Lett 2015;19(7):1185–8. https://doi.org/10.1109/
buffer size at WAP. The new protocol increases the overall throughput of LCOMM.2015.2424696.
[8] Song L, Liao Y, Bian K, Song L, Han Z. Cross-layer protocol design for csma/cd in
the network. In future this work can be extended for non-synchronous full-duplex wifi networks. IEEE Commun Lett 2016;20(4):792–5. https://doi.org/
transmissions. 10.1109/LCOMM.2016.2519518.
[9] Duarte M, Dick C, Sabharwal A. Experiment-driven characterization of full-duplex
wireless systems. IEEE Trans Wireless Commun 2012;11(12):4296–307. https://
Declaration of Competing Interest doi.org/10.1109/TWC.2012.102612.111278.
[10] Raut P, Sharma PK, Kothari A. Fd multi-user mobile system with timing errors.
The authors declare that they have no known competing financial IEEE Commun Lett 2019;23(12):2394–7. https://doi.org/10.1109/
LCOMM.2019.2939247.
interests or personal relationships that could have appeared to influence
[11] Gradshteyn IS, Ryzhik IM. Table of integrals, series, and products. 7th ed.
the work reported in this paper. Amsterdam: Elsevier/Academic Press; 2007.
[12] Bianchi G. Performance analysis of the ieee 802.11 distributed coordination
Acknowledgement function. IEEE J Sel Areas Commun 2000;18(3):535–47. https://doi.org/10.1109/
49.840210.
[13] Rappaport TS. Wireless communications: principles and practice. 1st ed. IEEE
This work was supported by the Early Career Research grant (ECR/ Press; 1996.