Document Version: Final Published version, also known as Publisher s PDF, Publisher s Final version or Version of Record

(1)

A MAC Protocol for Full Duplex Cellular Networks Chan, S.

Published in:

Procedia Computer Science Published: 01/05/2017

Document Version:

Final Published version, also known as Publisher’s PDF, Publisher’s Final version or Version of Record

License:

CC BY-NC-ND

Publication record in CityU Scholars:

Go to record

Published version (DOI):

10.1016/j.procs.2017.05.308

Publication details:

Chan, S. (2017). A MAC Protocol for Full Duplex Cellular Networks. In E. Shakshuki (Ed.), Procedia Computer Science (Vol. 109, pp. 140-147) https://doi.org/10.1016/j.procs.2017.05.308

Citing this paper

Please note that where the full-text provided on CityU Scholars is the Post-print version (also known as Accepted Author Manuscript, Peer-reviewed or Author Final version), it may differ from the Final Published version. When citing, ensure that you check and use the publisher's definitive version for pagination and other details.

General rights

Copyright for the publications made accessible via the CityU Scholars portal is retained by the author(s) and/or other copyright owners and it is a condition of accessing these publications that users recognise and abide by the legal

requirements associated with these rights. Users may not further distribute the material or use it for any profit-making activity or commercial gain.

Publisher permission

Permission for previously published items are in accordance with publisher's copyright policies sourced from the SHERPA RoMEO database. Links to full text versions (either Published or Post-print) are only available if corresponding publishers allow open access.

Take down policy

Contact [email protected] if you believe that this document breaches copyright and provide us with details. We will remove access to the work immediately and investigate your claim.

Download date: 21/03/2022

(2)

ScienceDirect

Available online at www.sciencedirect.com

Procedia Computer Science 109C (2017) 140–147

Peer-review under responsibility of the Conference Program Chairs.

10.1016/j.procs.2017.05.308

Procedia Computer Science 00 (2016) 000–000

www.elsevier.com/locate/procedia

The 8th International Conference on Ambient Systems, Networks and Technologies (ANT 2017)

A MAC Protocol for Full Duplex Cellular Networks

Sammy Chan^∗

Department of Electronic Engineering City University of Hong Kong Tat Chee Avenue, Kowloon Tong

Hong Kong

Abstract

Full duplex wireless communication is one of the enabling technologies of the fifth generation (5G) mobile telecommunication networks. It boosts up the bandwidth by enabling simultaneous data transmission and reception on a single channel. In this paper, we design a suitable medium access control (MAC) protocol for 5G networks to realize the benefit of full duplex communication.

We also develop an analytical model to evaluate the performance of our proposed MAC protocol.

2016 The Authors. Published by Elsevier B.V.c

Keywords: full duplex; MAC; cellular networks

1. Introduction

With the tremendous success of smart phones, mobile telecommunications networks need to support higher user density, and more bandwidth intensive and time-sensitive multimedia applications. In addition, the bandwidth de- mand is further increased by newly emerging technologies such as Internet of Things (IoT). These factors drive the evolution of the fourth-generation (4G) mobile telecommunications networks to the fifth generation (5G). Currently, 5G networks are under intensive research.

Compared with the 4G networks, 5G networks are expected to provide higher bandwidth by several orders of magnitude. This requires a multitude of new technologies, such as spectrum efficiency optimization, cooperative communications and multiple-input multiple-output (MIMO), to significantly enhance spectrum efficiency. Besides these new technologies, the recently proposed full duplex (FD) transmission techniques^1,2also offer great potential to significantly boosting spectrum efficiency.

Currently, frequency-division duplexing (FDD) and time-division duplexing (TDD) transmission techniques are commonly deployed. Both of them operate in half-duplex modes. In FDD, two frequency channels are needed to

∗Corresponding author. Tel.: +852-3442-7781 ; fax: +852-3442-0562.

E-mail address: [email protected]

1877-0509 c 2016 The Authors. Published by Elsevier B.V.

The 8th International Conference on Ambient Systems, Networks and Technologies (ANT 2017)

A MAC Protocol for Full Duplex Cellular Networks

Sammy Chan^∗

Hong Kong

Abstract

1. Introduction

∗Corresponding author. Tel.: +852-3442-7781 ; fax: +852-3442-0562.

(3)

Sammy Chan / Procedia Computer Science 109C (2017) 140–147 141 Available online at www.sciencedirect.com

The 8th International Conference on Ambient Systems, Networks and Technologies (ANT 2017)

A MAC Protocol for Full Duplex Cellular Networks

Sammy Chan^∗

Hong Kong

Abstract

1. Introduction

∗ Corresponding author. Tel.: +852-3442-7781 ; fax: +852-3442-0562.

The 8th International Conference on Ambient Systems, Networks and Technologies (ANT 2017)

A MAC Protocol for Full Duplex Cellular Networks

Sammy Chan^∗

Hong Kong

Abstract

1. Introduction

∗ Corresponding author. Tel.: +852-3442-7781 ; fax: +852-3442-0562.

2 Author name / Procedia Computer Science 00 (2016) 000–000

support bidirectional communications; one for uplink and one for downlink. In TDD mode, the uplink and downlink data are sent in orthogonal time-slots. On the other hand, FD transmission is the techniques which allow a node to simultaneously send and receive data on a single channel. According to the approach proposed by Choi et al.¹, two transmitting (TX) antennas and one receiving (RX) antenna are installed in each node. The RX antenna is positioned in a location such that the signals from the two TX antennas cancel each other. Thus, the transmitted signal does not interfere with the received signal and FD transmission is realized. However, this approach suffers from a bandwidth constraint and is not suitable for broadband wireless communications. In the approach proposed in², FD transmission is realized by the use of analog and digital cancellation. It requires one RX antenna, one TX antenna and a balanced/unbalunced (Balun) transformer. First, analog cancellation is used. The signal from the TX antenna is cancelled at the RX antenna by the inverted signal generated by the Balun transformer. It can cancel a minimum of 45 dB across a 40 MHz spectrum. Then, digital cancellation is employed to further reduce self-interference by up to 73 dB for 10 MHz OFDM signal.

When FD transmission is enabled, it can be deployed in two different modes. As shown in Fig. 1(a) , both nodes 1 and 2 are equipped with one TX antenna and one RX antenna. They can send and receive data simultaneously over the single channel. This mode is referred to as FD-bidirectional transmission. Alternatively, as shown in Fig. 1(b), node 1 is sending data to node 2. While node 2 is receiving the data, it can immediately forward it to node 3. In other words, node 2 is simultaneously receiving and transmitting. This mode is referred to as FD-relay transmission. For this mode, only node 2 needs to have two antennas installed; one TX antenna and one RX antenna. Since node 1 and node 3 do not transmit and receive simultaneously, each of them only need to have one antenna.

1 2 1 2 3

(a) (b)

Fig. 1. FD transmission modes.

Having FD transmission enabled at the physical layer, there is a need for a suitable medium access control (MAC) protocol to realize its benefit. So far, MAC protocols for FD transmission are mainly designed for wireless local area networks (WLANs). In this paper, we propose a MAC protocol for cellular networks in which both base stations and user equipments can support FD transmission. We also develop an analytical model to evaluate the average packet delay and its standard deviation under the saturated condition. The remainder of this paper is organized as follows. In Section 2, we review some recently proposed full duplex MAC protocols. In Section 3, we present our proposed MAC protocol. Then, we develop an analytical model in Section 4. Explicit expressions for the average packet delay, its standard deviation, and frame utilization are given. In Section 5, we verify the accuracy of our model by comparing with simulation results. We also investigate the impact of system parameters on the MAC layer performance. Finally, Section 6 concludes the paper.

2. Related Work

Besides proposing an innovative radio design to support full duplex operation on a single channel,²also suggests a MAC protocol to exploit the benefit of full duplex radios. It is based on the CSMA/CA protocol, each node accesses the channel by following the standard procedures of CSMA/CA. The difference comes when a node is transmitting.

Consider an example shown in Fig. 2, node 1 is sending a packet to node 2 after carrier sensing. Once node 2 has received the header information, it realizes that node 1 is sending a packet to itself. Since full duplex is allowed in a single channel, node 2 can also immediately start transmitting a packet to node 1. Note that this MAC protocol naturally mitigates the hidden terminal problem. Referring again to Fig. 2, node 3 is a hidden terminal to Node 1 and could cause collision at node 2. However, with this MAC protocol, transmission at node 1 also triggers simultaneous transmission at node 2. node 3 thus could sense that the channel is busy and will not initiate any transmission, which leads to collision. This works well if both nodes finish transmission at the same time. However, if node 2 has no packet for node 1, or transmits a smaller packet and hence finishes transmission before node 1, node 3 can still be a

(4)

142 Sammy Chan / Procedia Computer Science 109C (2017) 140–147

Author name / Procedia Computer Science 00 (2016) 000–000 3

hidden terminal and cause collision. This problem is resolved by requiring node 2 to transmit a busy tone after it has finished transmitting the packet, until node 1 finishes transmission.

1

header

2 3 TX12

TX21

TX12

TX21

suppress hidden node

busytone

Fig. 2. Full duplex operation in the MAC of².

In³, a MAC protocol is proposed for nodes equipped with directional antennas to improve the throughput of a multi-hop flow. Let us first consider the usual case that omni-directional antennas are used in nodes with full duplex capability. For the multi-hop flow from node 1 to node 4 as shown in Figure 3, at any time, either both node 1 and node 2 are simultaneously transmitting or only node 3 is transmitting. This is because if all of nodes 1,2 and 3 are transmitting at the same time, node 3 would cause interference at node 2. Due to this constraint, the end-to-end throughput is 1/2, assuming that the bandwidth of each hop is 1. Now, if the omni-directional antenna in each node is replaced by two 180^odirectional antennas, and only one of them is activated at ay time, as suggested in³, interference at node 2 due to node 3 is eliminated. This allows nodes 1, 2 and 3 to transmit simultaneously. Hence, the end-to-end throughput is increased to 1. The MAC protocol accompanying this node architecture is based on CSMA/CA, but with three modifications. First, a node does not need to always stay silent when it senses carrier. Instead, if a node detects that the destination MAC address of the packet being transmitted on the channel is its own MAC address, it is allowed to transmit data to another node. This modification enables nodes 1, 2 and 3 to transmit simultaneously. Second, upon receiving a packet, a node does not need to return any acknowledgment packet. As a result, collisions between data and acknowledgment packets can be avoided. Third, since the above two modifications effectively eliminate the possibility of collision, the backoff procedure in CSMA/CA is removed.

1 2 3 4

Fig. 3. A multi-hop flow from node 1 to node 4.

In a WLAN, when the traffic between clients and AP is asymmetric, the opportunity for full duplex operation decreases. Moreover, since AP needs to contend with all clients, its transmission queue could build up. The MAC protocol of² is enhanced by dynamically changing the contention window of AP to balance uplink and downlink traffic⁴. According to the MAC protocol proposed in⁴, after AP has finished a data frame transmission, it checks the length its transmission queue. If the queue length is below a pre-defined threshold value, its contention window is set to CWl. Otherwise, the contention window is set to CWs, where CWl>CWs. This essentially gives AP a higher priority to access the channel when its traffic load is high.

In⁵, a unified MAC protocol is proposed to establish either FD-bidirectional or FD-relay transmission. The protocol is based on the RTS/CTS mechanism. Referring to Fig. 3, when node 1 has a packet to send, it will first send a RTS packet, which contains the source and destination addresses. After node 2 has received the RTS packet, it checks the destination address contained in the packet. If the destination address is its own address, node 2 sends a CTS packet with source and destination addresses set to node 2 and node 1, respectively. Then, node 1 will send another CTS to complete the three-way handshake, and subsequently, both nodes 1 and 2 start the FD-bidirectional transmission. On the other hand, if the destination address contained in the RTS packet is set to node 3, node 2 will send a CTS packet with source and destination addresses set to node 2 and node 3, respectively. Upon receiving this CTS packet, node 3 sends a CTS packet to complete the three-way handshake, and subsequently, both nodes 1 and 2 start the FD-relay transmission.

(5)

Sammy Chan / Procedia Computer Science 109C (2017) 140–147 143

hidden terminal and cause collision. This problem is resolved by requiring node 2 to transmit a busy tone after it has finished transmitting the packet, until node 1 finishes transmission.

1

header

2 3 TX12

TX21

TX12

TX21

suppress hidden node

busytone

Fig. 2. Full duplex operation in the MAC of².

In³, a MAC protocol is proposed for nodes equipped with directional antennas to improve the throughput of a multi-hop flow. Let us first consider the usual case that omni-directional antennas are used in nodes with full duplex capability. For the multi-hop flow from node 1 to node 4 as shown in Figure 3, at any time, either both node 1 and node 2 are simultaneously transmitting or only node 3 is transmitting. This is because if all of nodes 1,2 and 3 are transmitting at the same time, node 3 would cause interference at node 2. Due to this constraint, the end-to-end throughput is 1/2, assuming that the bandwidth of each hop is 1. Now, if the omni-directional antenna in each node is replaced by two 180^odirectional antennas, and only one of them is activated at ay time, as suggested in³, interference at node 2 due to node 3 is eliminated. This allows nodes 1, 2 and 3 to transmit simultaneously. Hence, the end-to-end throughput is increased to 1. The MAC protocol accompanying this node architecture is based on CSMA/CA, but with three modifications. First, a node does not need to always stay silent when it senses carrier. Instead, if a node detects that the destination MAC address of the packet being transmitted on the channel is its own MAC address, it is allowed to transmit data to another node. This modification enables nodes 1, 2 and 3 to transmit simultaneously. Second, upon receiving a packet, a node does not need to return any acknowledgment packet. As a result, collisions between data and acknowledgment packets can be avoided. Third, since the above two modifications effectively eliminate the possibility of collision, the backoff procedure in CSMA/CA is removed.

1 2 3 4

Fig. 3. A multi-hop flow from node 1 to node 4.

In a WLAN, when the traffic between clients and AP is asymmetric, the opportunity for full duplex operation decreases. Moreover, since AP needs to contend with all clients, its transmission queue could build up. The MAC protocol of² is enhanced by dynamically changing the contention window of AP to balance uplink and downlink traffic⁴. According to the MAC protocol proposed in⁴, after AP has finished a data frame transmission, it checks the length its transmission queue. If the queue length is below a pre-defined threshold value, its contention window is set to CWl. Otherwise, the contention window is set to CWs, where CWl >CWs. This essentially gives AP a higher priority to access the channel when its traffic load is high.

In⁵, a unified MAC protocol is proposed to establish either FD-bidirectional or FD-relay transmission. The protocol is based on the RTS/CTS mechanism. Referring to Fig. 3, when node 1 has a packet to send, it will first send a RTS packet, which contains the source and destination addresses. After node 2 has received the RTS packet, it checks the destination address contained in the packet. If the destination address is its own address, node 2 sends a CTS packet with source and destination addresses set to node 2 and node 1, respectively. Then, node 1 will send another CTS to complete the three-way handshake, and subsequently, both nodes 1 and 2 start the FD-bidirectional transmission. On the other hand, if the destination address contained in the RTS packet is set to node 3, node 2 will send a CTS packet with source and destination addresses set to node 2 and node 3, respectively. Upon receiving this CTS packet, node 3 sends a CTS packet to complete the three-way handshake, and subsequently, both nodes 1 and 2 start the FD-relay transmission.

In⁶, the authors propose a MAC protocol called Synchronized Contention Window Full Duplex (S-CW FD) to support full duplex operation in a WLAN. S-CW FD is a modified form of IEEE 802.11 DCF protocol⁷. Its basic idea is to synchronize two FD nodes by sharing the information about the size of the next backoff window so that they can access the channel at the same time and carry out FD transmission. In order to start FD-bidirectional transmission in S-CW FD, a node (either a station or the AP) first needs to achieve a successful packet transmission based on DCF.

This packet coveys three control fields: fd is a one-bit field to inform the receiving node to prepare for FD-bidirectional transmission, next bo contains the number of backoff slots of the sending node for the next contention period, and fd-master is a one-bit field determining whether the receiving node will act as a master or a slave. If the receiving node acts as a slave, it will adopt next bo as the next backoff value. After the transmission of the current packet is completed and acknowledged, the sending and receiving nodes will carry out the backoff, and then start the full duplex transmission. The advantages of S-CW FD is that it is compatible with legacy nodes operating the traditional DCF protocol, and that it can operate in both infrastructure and ad hoc modes.

3. MAC protocol

The frame structure for the proposed MAC protocol is shown in Fig. 4. In this protocol, the channel is time slotted and organized into frames. Each frame has a duration of E, and consists of a request subframe, a full duplex data subframe, an information subframe and a downlink data subframe, with duration Tr, Tf, Tiand Th, respectively.

The base station (BS) can control the duration of the request subframe, full duplex data subframe and downlink data subframe by broadcasting the values of Tr, Tfand Tiin the information subframe.

Tf

Th

Tr

frame 1

V

BW request



packet sent

MapDL FD Map

U

full duplex

full duplex full duplex

Ti

frame 2 frame 3

Tr Tr

Fig. 4. Frame structure of the proposed MAC protocol.

A request subframe has m request slots, each of length t. In order to gain the right to send a data packet, an user equipment (UE) first needs to send a bandwidth request to a randomly chosen slot in the request subframe. If there is only one request submitted to the request slot, the request is successful. On the other hand, if there are two or more UE submitting their requests in the same request slot, collision occurs. For simplicity, in this paper, we assume that the BS also allocates m data slots in a full duplex data subframe. Upon receiving the bandwidth requests, the BS will announce the contention result in the full duplex map (FD Map) of the information subframe. It assigns a data slot in the full duplex data subframe of the next frame to each successful UE. Each data slot is of length T (T t), which is the transmission time of a data packet (all packets are assumed to have the same length throughout this paper).

Here, we assume that a data slot is randomly chosen and assigned to a successful UE. Since full duplex transmission is supported, when an UE is transmitting a packet to the BS in the assigned data slot, the BS can simultaneously send downlink traffic to the corresponding UE. For those unsuccessful UE, they need to execute a certain backoff algorithm to determine when to retransmit the request. Here, for simplicity, we assume the UE are persistent. That is, they will send requests again in the immediately following frame. In Fig. 4, we show a particular example in which an UE sends a request in frame 1 but fails. It re-attempts in frame 2 and succeeds, so it can have a duplex data transmission in frame 3. If the traffic between UE and the BS is symmetric, this MAC protocol is sufficient. However, to cater for the

(6)

case in which some UE only have downlink traffic, the downlink data subframe with b data slots is provisioned. When an UE only has downlink traffic, the BS will inform it in the downlink map (DL Map) of the information subframe to receive a packet in a particular data slot in the downlink data subframe.

4. Analytical Model

Consider a cell with N UE. Assume that all UE are in saturated condition, i.e., they always have packets to send.

The packet delay X is defined as the time duration from its first bandwidth request until the packet transmission has finished. Note that if the bandwidth reservation of a packet is successful, the packet will be removed from the head of the queue into a temporary buffer and transmitted to the channel in the coming full duplex data subframe. Thus the bandwidth request of the next packet (now become the head of the queue) can be sent in the request subframe of the next frame.

Given that there are N saturated UE, as discussed in the previous section, an active UE will randomly choose a request slot among m available slots to send its bandwidth request. The probability that the request of a tagged UE will be successful is given by

p =m 1

1 m

1 − 1

m

_N−1

=

1 − 1

m

_N−1

. (1)

Thus, with probability p given in (1), the tagged UE will be allocated a data slot by the BS in the subsequent frame to transmit a packet. Referring to Fig. 4, let U and V be the random variables (RVs) representing the time durations from the time of sending the request until the end of the request subframe, and from the beginning of the appropriate full duplex data subframe until a packet of the tagged UE has been transmitted, respectively. Clearly, Tr=mt, Tf =mT and E = Tr+Tf+Ti+Th.

Recall that upon a successful request, an UE is only allocated a data slot in the full duplex data subframe of the subsequent frame. Therefore, given that the tagged UE is successful in its first attempt of sending bandwidth request, referring to Fig. 4, X is given by

X = U + E + V w.p. p, where w.p. stands for “with probability”.

For the case when the tagged UE fails in its first attempt but succeeds in the second attempt of sending bandwidth request, the packet delay can be expressed as

X = U + 2E + V w.p. p(1 − p),

where p(1 − p) is the probability that the request is successful in the second attempt. In summary, we give the probability mass function (pmf) of X as in (2) assuming the tagged UE will keep sending bandwidth request until it is successful.

X =









U + E + V w.p. p, U + 2E + V w.p. p(1 − p), U + 3E + V w.p. p(1 − p)², ...

(2)

To complete the expression of X, we now determine the pmf of U and V. As the tagged UE randomly chooses the request slot, the pmf of the U is then given below

U =









mt w.p. 1/m,

(m − 1)t w.p. 1/m, ...

t w.p. 1/m.

(3)

Equation (3) assumes that an UE always sends its request at the beginning of the chosen slot.

(7)

case in which some UE only have downlink traffic, the downlink data subframe with b data slots is provisioned. When an UE only has downlink traffic, the BS will inform it in the downlink map (DL Map) of the information subframe to receive a packet in a particular data slot in the downlink data subframe.

4. Analytical Model

Consider a cell with N UE. Assume that all UE are in saturated condition, i.e., they always have packets to send.

The packet delay X is defined as the time duration from its first bandwidth request until the packet transmission has finished. Note that if the bandwidth reservation of a packet is successful, the packet will be removed from the head of the queue into a temporary buffer and transmitted to the channel in the coming full duplex data subframe. Thus the bandwidth request of the next packet (now become the head of the queue) can be sent in the request subframe of the next frame.

Given that there are N saturated UE, as discussed in the previous section, an active UE will randomly choose a request slot among m available slots to send its bandwidth request. The probability that the request of a tagged UE will be successful is given by

p =m 1

1 m

1 − 1

m

_N−1

=

1 − 1

m

_N−1

. (1)

Thus, with probability p given in (1), the tagged UE will be allocated a data slot by the BS in the subsequent frame to transmit a packet. Referring to Fig. 4, let U and V be the random variables (RVs) representing the time durations from the time of sending the request until the end of the request subframe, and from the beginning of the appropriate full duplex data subframe until a packet of the tagged UE has been transmitted, respectively. Clearly, Tr=mt, Tf =mT and E = Tr+Tf +Ti+Th.

Recall that upon a successful request, an UE is only allocated a data slot in the full duplex data subframe of the subsequent frame. Therefore, given that the tagged UE is successful in its first attempt of sending bandwidth request, referring to Fig. 4, X is given by

X = U + E + V w.p. p, where w.p. stands for “with probability”.

For the case when the tagged UE fails in its first attempt but succeeds in the second attempt of sending bandwidth request, the packet delay can be expressed as

X = U + 2E + V w.p. p(1 − p),

where p(1 − p) is the probability that the request is successful in the second attempt. In summary, we give the probability mass function (pmf) of X as in (2) assuming the tagged UE will keep sending bandwidth request until it is successful.

X =









U + E + V w.p. p, U + 2E + V w.p. p(1 − p), U + 3E + V w.p. p(1 − p)², ...

(2)

To complete the expression of X, we now determine the pmf of U and V. As the tagged UE randomly chooses the request slot, the pmf of the U is then given below

U =









mt w.p. 1/m,

(m − 1)t w.p. 1/m, ...

t w.p. 1/m.

(3)

Equation (3) assumes that an UE always sends its request at the beginning of the chosen slot.

Next, we consider how to obtain the pmf of V. Let k be the maximum possible number of successful requests among all the UE in the frame immediately before the frame where the packet of the tagged UE is transmitted (clearly the tagged node is among those who successfully transmitted their requests). Since there are m request slots and N UE, we have k = min(m, N).

Since a request slot is chosen randomly, the probability that an UE attempts to send the request in a slot is given by τ =1/m.

The probability that there are j, 0 ≤ j ≤ k = min(m, N) successful requests among m request slots can be approximated based on a truncated binomial distribution

Q( j) =

_m

j

ξ^j(1 − ξ)^{m− j}

k i=0

m i

ξⁱ(1 − ξ)^m−i

, (4)

where ξ = Nτ(1 − τ)^N−1is the probability that a request sent in a request slot will be successful given that there are N UE each attempting to send requests with probability τ.

The probability that there are j successful requests other than the tagged UE is given by

q( j) = Q( j + 1)/(1 − Q(0)), 0 ≤ j ≤ k − 1. (5) Since the BS randomly allocates data slots among successful requests, using (5) the pmf of V can be expressed as

V =











T w.p. ^k−1

j=0

1 j + 1q( j), 2T w.p. ^k−1

j=1

1 j + 1q( j), ...

kT w.p. ¹_k q(k − 1).

(6)

Next, we express the mean and second moment of the packet delay (X, X²) using the means and second moments of their constituent random variables U, V. Unless otherwise specified, we will denote the expectation operator of a RV Z by Z in the rest of this paper.

From (2) we write X, X²as follows X = U + V + E

p, (7)

X²=U²+V²+2U V +2

p E(U + V) +2 − p p² E².

Note that X²is the second moment of X and is calculated by squaring and taking expected value of both sides of (2).

It remains to determine U, U², V and V²from (3) and (6), and can be expressed as

U = (m + 1)t/2, (8)

U² =(m + 1)(2m + 1)t²/6, V = T^k−1

j=0

q( j)

j i=0

i + 1 j + 1,

V² =T²

k−1 j=0

q( j)

j i=0

(i + 1)² j + 1 ,

(8)

where q( j) is given in (5).

To measure the efficiency of the MAC protocol in terms of bandwidth usage, we define the utilization of the full duplex data subframe as the average ratio between the number of occupied data slots and the total number of available data slots. Using (4) the utilization η of full duplex data subframe can be expressed as η = ^k_j=0 ^{jQ( j)}_m .

5. Numerical Results

In this section, we use the developed analytical model to investigate the impact of different parameters on the performance of the MAC protocol. At the same time, we also validate the analytical model by simulation. For this purpose, we have built a simulator to generate simulation results. The simulator is written in C++ and based on the discrete event-driven simulation technique. The duration of each simulation run is 5,000 seconds, with a warm-up period of 500 seconds.

The system parameters used in the following performance evaluation of the MAC protocol are as follows. With a 25 MHz spectrum, the 64-QAM modulation scheme is used to achieve a data rate of 120 Mbps. A mini slot is a basic unit of different time slots, and has a duration of ₂₅₀₀¹ millisecond. Each bandwidth request has a duration of 6 mini slots. Each data slot has a duration of 94 mini slots, which allows the transmission of approximately 0.5 KB data at 120 Mbps. The information subframe has a duration of 10 request slots.

First, we set b = 12 and m = 15, and evaluate the delay and utilization of the MAC protocol for different N.

The results are shown in Fig. 5(a) and Fig. 5(b), respectively. From these figures (and later ones as well), it can be seen that the analytical results match well with their corresponding values from the simulation. This demonstrates that our analytical model is accurate. Fig. 5(a) shows that for a given number of bandwidth request slots, the mean and standard deviation of packet delay increase exponentially as the number of UE increases. This is because as N increases, collisions happen more frequently and an UE has to make more attempts for its request to be successful.

Hence the delay also increases. For utilization, as shown in Fig. 5(b), it first increases with N up to a maximum value.

This is the light load region, a larger N means more requests are made and it results in more successful requests.

When N increases further, the load becomes heavier and there are more collisions and unsuccessful requests, hence the utilization decreases.

10 15 20 25

0 1 2 3 4 5 6

Number of user equipment (N)

Delay (ms)

analysis mean simulation mean analysis stdev simulation stdev

(a)

10 15 20 25

0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4

Utilization

analysis simulation

(b)

Fig. 5. (a) Mean and standard deviation of delay and (b) utilization versus the number of user equipments.

Next, we set b = 10 and N = 15, and evaluate the delay and utilization of the MAC protocol for different m. The results are shown in Fig. 6(a) and Fig. 6(b), respectively. Fig. 6(a) shows that when the number of UE is fixed, the mean and standard deviation of packet delay decrease exponentially as the number of request slots increases. This is because as m increases, the success probability of requests also increases. Hence, less re-attempts are needed for a request to be successful. For utilization as shown in 6(b), the trend is similar to that of Fig. 5(b). When m is small, increase in m reduces collision probability and increases utilization. Once the utilization has reached the maximum value, further increase in m only means the request slots are over provisioned, hence the utilization decreases.

Next,we investigate the impact of the size of the downlink data subframe. We consider that the traffic is close to symmetric and, only occasionally, the BS has to rely on the downlink data subframe to deliver packets to UE. In that case, the number of data slots in this subframe can be much smaller than m. Here, we use the case of b = 12, m = 15, N = 20 as a reference, and calculate how much (measured in %) delay is reduced when b is reduced from 12 to

(9)

where q( j) is given in (5).

To measure the efficiency of the MAC protocol in terms of bandwidth usage, we define the utilization of the full duplex data subframe as the average ratio between the number of occupied data slots and the total number of available data slots. Using (4) the utilization η of full duplex data subframe can be expressed as η = ^k_j=0 ^{jQ( j)}_m .

5. Numerical Results

In this section, we use the developed analytical model to investigate the impact of different parameters on the performance of the MAC protocol. At the same time, we also validate the analytical model by simulation. For this purpose, we have built a simulator to generate simulation results. The simulator is written in C++ and based on the discrete event-driven simulation technique. The duration of each simulation run is 5,000 seconds, with a warm-up period of 500 seconds.

The system parameters used in the following performance evaluation of the MAC protocol are as follows. With a 25 MHz spectrum, the 64-QAM modulation scheme is used to achieve a data rate of 120 Mbps. A mini slot is a basic unit of different time slots, and has a duration of ₂₅₀₀¹ millisecond. Each bandwidth request has a duration of 6 mini slots. Each data slot has a duration of 94 mini slots, which allows the transmission of approximately 0.5 KB data at 120 Mbps. The information subframe has a duration of 10 request slots.

First, we set b = 12 and m = 15, and evaluate the delay and utilization of the MAC protocol for different N.

The results are shown in Fig. 5(a) and Fig. 5(b), respectively. From these figures (and later ones as well), it can be seen that the analytical results match well with their corresponding values from the simulation. This demonstrates that our analytical model is accurate. Fig. 5(a) shows that for a given number of bandwidth request slots, the mean and standard deviation of packet delay increase exponentially as the number of UE increases. This is because as N increases, collisions happen more frequently and an UE has to make more attempts for its request to be successful.

Hence the delay also increases. For utilization, as shown in Fig. 5(b), it first increases with N up to a maximum value.

This is the light load region, a larger N means more requests are made and it results in more successful requests.

When N increases further, the load becomes heavier and there are more collisions and unsuccessful requests, hence the utilization decreases.

10 15 20 25

0 1 2 3 4 5 6

Delay (ms)

(a)

10 15 20 25

0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4

Utilization

analysis simulation

(b)

Fig. 5. (a) Mean and standard deviation of delay and (b) utilization versus the number of user equipments.

Next, we set b = 10 and N = 15, and evaluate the delay and utilization of the MAC protocol for different m. The results are shown in Fig. 6(a) and Fig. 6(b), respectively. Fig. 6(a) shows that when the number of UE is fixed, the mean and standard deviation of packet delay decrease exponentially as the number of request slots increases. This is because as m increases, the success probability of requests also increases. Hence, less re-attempts are needed for a request to be successful. For utilization as shown in 6(b), the trend is similar to that of Fig. 5(b). When m is small, increase in m reduces collision probability and increases utilization. Once the utilization has reached the maximum value, further increase in m only means the request slots are over provisioned, hence the utilization decreases.

Next,we investigate the impact of the size of the downlink data subframe. We consider that the traffic is close to symmetric and, only occasionally, the BS has to rely on the downlink data subframe to deliver packets to UE. In that case, the number of data slots in this subframe can be much smaller than m. Here, we use the case of b = 12, m = 15, N = 20 as a reference, and calculate how much (measured in %) delay is reduced when b is reduced from 12 to

10 11 12 13 14 15 16 17 18 19 20

1 1.5 2 2.5 3 3.5 4

Number of request slots (m)

Delay (ms)

(a)

10 11 12 13 14 15 16 17 18 19 20

0.3 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.4

Number of request slots (m)

Utilization

analysis simulation

(b)

Fig. 6. (a) Mean and standard deviation of delay and (b) utilization versus the number of request slots.

2. The results are plotted in Fig. 7. It can seen that when b = 2, the mean delay is reduced by 35%. Clearly, the reduction of delay is due to a smaller frame size as th decreases with b. This reflects one of the benefits of the full duplex MAC protocol.

2 3 4 5 6 7 8 9 10 11 12

0 5 10 15 20 25 30 35 40

Number of slots in the downlink data subframe.

time reduction (%)

Fig. 7. The percentage of delay reduction versus the number of slots in downlink data subframe.

6. Conclusion

In this paper, we have proposed a MAC protocol for mobile telecommunication networks which support full duplex wireless communication. We have also developed an analytical model for performance evaluation in terms of mean and standard deviation of packet delay, and utilization of the full duplex data subframe. Explicit forms of these performance metrics have been derived and validated against simulations. From the numerical results, we have shown that suitable system parameters can be chosen to maximize the utilization. Moreover, packet delay can be significantly reduced due to the deployment of full duplex wireless communication.

References

1. Choi J, Jain M, Choi J, Srinivasan K, Levis P, Katti S. Achieving single channel, full duplex wireless communication. Proceedings, ACM MobiCom10, Chicago, 20-24 Sep. 2010, pp. 1-12.

2. Jain M, Choi JI, Kim TM, Bharadia D, Seth S, Srinivasan K, Levis P, Katti S, Sinha P. Practical, real-time, full duplex wireless. Proceedings, ACM MobiCom’11, Las Vegas, 19-23 Sep. 2011, pp. 300–312.

3. Miura K, Bandai M. Node Architecture and MAC Protocol for Full Duplex Wireless and Directional Antennas. Proceedings, IEEE International Symposium on Personal, Indoor and Mobile Radio Communications (PIMRC’12), Sydney, 9-12 Sep. 2012, pp. 385–390.

4. Oashi S, Bandai M. Performance of Medium Access Control Protocols for Full-Duplex Wireless LANs. Proceedings, Asia-Pacific Symposium on Information and Telecommunication Technologies (APSITT’12), Santiago and Valparaiso, Republic of Chile, 5-9 Nov. 2012.

5. Cheng W, Zhang X, Zhang H. RTS/FCTS Mechanism Based Full-duplex Mac Protocol for Wireless Networks. Proceedings, IEEE Globecom 2013, Alanta, 9-13 Dec. 2013, pp. 5017–5022.

6. Marlah D, Gürbüz Ö. S-CW FD: A MAC Protocol for Full-duplex in Wireless Local Area Networks. Proceedings, IEEE Wireless Communica- tions and Networking Conference (WCNC 2016), San Francisco, 3-6 Mar. 2016.

7. IEEE standard for information technologytelecommunications and information exchange between systems local and metropolitan area net- worksspecific requirements part 11: Wireless lan medium access control (mac) and physical layer (phy) specifications. IEEE Std 802.11- 2012 (Revision of IEEE Std 802.11-2007), March 2012, pp. 1–2793.