Millimeter-Wave Wireless LAN and Its Extension Toward 5G Heterogeneous Networks
Millimeter-Wave Wireless LAN and Its Extension Toward 5G Heterogeneous Networks
Millimeter-Wave Wireless LAN and Its Extension Toward 5G Heterogeneous Networks
SUMMARY Millimeter-wave (mmw) frequency bands, especially 60 GHz introduced to address this issue. The main focus of this paper
unlicensed band, are considered as a promising solution for gigabit short is on HetNets using coordinated mmw WLANs. Such an
range wireless communication systems. IEEE standard 802.11ad, also
known as WiGig, is standardized for the usage of the 60 GHz unlicensed architecture not only takes the advantage of HetNet
band for wireless local area networks (WLANs). By using this mmw architecture but also benefits from huge available bandwidth
WLAN, multi-Gbps rate can be achieved to support bandwidth-intensive in mmw band. This technique will be studied throughout this
multimedia applications. Exhaustive search along with beamforming (BF) paper as a potential enabler for future 5G Networks.
is usually used to overcome 60 GHz channel propagation loss and
Latest advances in mmw antennas and packaging
accomplish data transmissions in such mmw WLANs. Because of its short
range transmission with a high susceptibility to path blocking, multiple technology [5] allow creating phased antenna arrays with
number of mmw access points (APs) should be used to fully cover a typical limited number of elements. Next evolution in mmw
target environment for future high capacity multi-Gbps WLANs. Therefore, technology is analog/digital hybrid beamforming (BF), like
coordination among mmw APs is highly needed to overcome packet modular antenna arrays (MAA) [6] [7], consisting of a large
collisions resulting from un-coordinated exhaustive search BF and to
increase the total capacity of mmw WLANs. In this paper, we firstly give number of sub-array modules. Each module has a built-in
the current status of mmw WLANs with our developed WiGig AP prototype. sub-array phase control and coarse beam steering capability,
Then, we highlight the great need for coordinated transmissions among and fine BF is realized by digital signal processing. Such a
mmw APs as a key enabler for future high capacity mmw WLANs. Two high gain BF antenna has the potential to increase the range
different types of coordinated mmw WLAN architecture are introduced.
of mmw communication up to several hundreds of meters
One is the distributed antenna type architecture to realize centralized
coordination, while the other is an autonomous coordination with the for line-of-sight (LOS) backhaul and several tens of meters
assistance of legacy Wi-Fi signaling. Moreover, two heterogeneous for access [7] [8]. For communications in 60 GHz band, the
network (HetNet) architectures are also introduced to efficiently extend the ITU-R’s recommended channelization comprises four
coordinated mmw WLANs to be used for future 5th Generation (5G) cellular consecutive channels, each 2.16 GHz wide, centered at
networks.
58.32, 60.48, 62.64 and 64.80 GHz respectively, which
Keywords: millimeter wave, IEEE802.11ad, coordinated mmw WLAN, 5G allows a single RF device to operate worldwide [9] - [17].
cellular networks, heterogeneous networks. WiGig [18] and IEEE 802.11ad [19] are standards for the
usage of 60 GHz band using single carrier (SC) and OFDM
1. Introduction modulations with a maximum data rate of 6.7 Gbps [19].
Also, IEEE 802.11ad defines multi-band (2.4, 5 and 60
The exploding demand for the data traffic in cellular GHz) medium access control (MAC) protocol for backward
networks seems to be a persistent problem. According to [1], compatibility with legacy IEEE 802.11 a, b, g, and n (Wi-Fi)
in 2014 almost half a billion (497 million) mobile devices standards [20] and for fast session transfer (FST) between
and connections were added and the global mobile data them [19]. Antenna beamforming is usually used to combat
traffic reached 2.5 exabytes per month, which shows 69 60 GHz channel propagation losses and accomplish data
percent growth compared to 2013. This trend motivates a transmissions. Current IEEE802.11ad standard utilizes
constant effort for implementing more efficient networks exhaustive search BF protocols on the MAC layer, using
with higher capacity. Techniques such as massive MIMO [2], switched antenna array with a structured codebook, as a
heterogeneous networks (HetNets) using small cells [3], and suitable BF mechanism for the 60 GHz band [18] [19] [21]
bandwidth expansion using mmw technology [4] are all [22]. As prototyping activities, in Oct. 2014, Qualcomm
shows off WiGig with prototype tablets on Snapdragon 810
processors [23]. Also, Panasonic demonstrates a WiGig
† The author is with Osaka University, Osaka, Japan.
†† The author is with Aswan University, Aswan, Egypt.
access point (AP) in early 2015, which will be described in
††† The author is with Tokyo Institute of Technology, Tokyo more details in this paper.
Japan. All mmw architectures so far consider only point-to-point
†††† The author is with Panasonic Corporation, Japan. and point-to-multi-point applications [18] [19]. For future
††††† The author is with KDDI R&D Laboratories, Inc., Japan. high capacity mmw WLANs, multi-point-to-multi-point
transmissions, namely multiple 60GHz APs to cover
2
multiple terminals, should be used to fully cover a typical for LTE/mmw WLAN internetworking. Pros and cons of
target environment [24] - [30]. This architecture also each architecture candidate are explained in details in this
provides the potential to enhance the robustness against paper.
shadowing and achieves higher spatial spectral efficiency by The rest of this paper is organized as follows. Section 2
reducing the frequency reuse distance thanks to highly presents an overview of the current WiGig/IEEE802.11ad
directional characteristics of 60 GHz frequency band. based WLAN and also describes a possible extension toward
However, installing a numerous number of autonomously multiple mmw AP coordination. Section 3 provides the sub-
operated APs using CSMA/CA random access brings about cloud mmw WLAN where Wi-Fi and WiGig coordinate
several challenges related to the exhaustive search BF. For tightly via dual-band MAC protocol based on statistical
example, nearby APs may start the BF simultaneously, or an learning. Further extension toward 5G HetNet using mmw
AP may perform the exhaustive search BF when its nearby WLANs are given in Sect. 4. Finally, Sect. 5 concludes this
APs are involved in data transmissions. Both cases cause a paper. Table 1 summarizes important abbreviations used in
lot of packet collisions and degradation in the total system this paper.
performance due to hidden nodes if the APs are
autonomously operated. Therefore, a new system design is Table 1 Abbreviations.
Acronym Description
highly needed to coordinate mmw AP transmissions for 5G 5th generation
future multi-Gbps WLANs. To the best of our knowledge, A-BFT association BF training
ACK acknowledgement
coordination of mmw APs for concurrent transmissions in ANDSF access networks discovery and selection function
APC AP controller
random access scenario is never touched before in the API AP information
literature, while the authors in [31] proposed a coordinated ATI announcement transmission interval
BBU base band unit
beamforming assuming perfectly synchronous network. In BF beamforming
this paper, two types of coordinated mmw WLAN BHI beacon header interval
BID best beam identification
architecture are presented, which we proposed in [26]–[30]. BLI blocking information
The MAC protocols required to operate the coordinated BRP beam refinement protocol
BS base station
mmw WLANs are also given. The first mmw WLAN BT beam tracking
architecture considers the mmw APs as remote radio heads BTI beacon transmission interval
C/U control-plane/user-plane
(RRHs) connected to a WiGig AP base band unit (BBU), CAPEX capital expenditure
CBAP contention-based access period
which coordinates the BF and data transmissions operations CEF channel estimation field
among the installed APs in a full centralized manner [29] CLI candidate link information
CMOS complementary metal-oxide-semiconductor
[30]. The second architecture is a sub-cloud mmw WLAN, CSMA/CA carrier sense multiple access/collision avoidance
which utilizes the wide-coverage 5GHz (Wi-Fi) signal to CTS clear to send
DB database
autonomously coordinate the transmissions among dual- DBPSK differential BPSK
band (5, 60 GHz) mmw APs using a novel dual-band MAC DTI data transfer interval
E-ANDSF enhanced ANDSF
protocol based on statistical learning [26]- [28]. EIRP equivalent isotropic radiated power
In this paper, we also study the efficient interworking EPC evolved packet core
ePDG evolved packet data gateway
between mmw WLAN and cellular networks as a 5G enabler. E-UTRAN evolved universal terrestrial RAN
FBK feedback
Coordinated mmw WLANs can provide an extraordinary FST fast session transfer
high data rate which can be efficiently distributed in hotspot HDMI high definition multimedia interface
HetNet heterogeneous network
areas. However, due to high propagation loss and IC integrated circuit
penetration loss, the coverage area of mmw WLANs will be IRX inter-RAN exchange
LDO low drop-out regulator
in the range of several tens of meters [7] [8]. This issue LOS line-of-sight
makes it much more difficult for a user equipment (UE) to LP learning point
MAA modular antenna array
discover mmw small WLANs and handover to them. Here, MAC media access control
the general concepts of HetNets [32]-[34] and Control- MCS modulation and coding scheme
mmw millimeter-wave
plane/User-plane splitting (C/U splitting) [35] are exploited MNO mobile network operator
NAV network allocation vector
to facilitate the LTE/mmw WLAN internetworking. In this NF noise figure
paper, three HetNet architectures, including 3rd generation OFDM orthogonal frequency division multiplexing
OPEX operating expense
partnership project (3GPP) standard [36], are introduced as PAPR peak to average power ratio
candidates of LTE/mmw WLANs internetworking. In PDN packet data network
PHY Physical
contrast to the loosely coupled LTE/WLAN internetworking PMU power management unit
in 3GPP using access network discovery and selection PS phase shifter
RAN radio access network
function (ANDSF) [37], a tightly coupled LTE/WLAN RRH remote radio head
internetworking, proposed by the authors in [38], is RSS radio signal strength
RSSI RSS indicator
presented. In addition, a multi-layer LTE/WLAN/WiGig RTS request to send
SC single carrier
internetworking based on the concept of inter-RAN SIFS short inter frame space
exchange (IRX) with enhanced ANDSF (E-ANDSF), SLS sector-level sweep
SP service period
proposed by the authors in [39], is also given as a candidate
3
STA Station
STF short training field
UE user equipment
US, Canada and Korea
USB universal serial bus
WLAN wireless local area network
EU and Japan
Australia
2. WiGig/IEEE 802.11ad based WLAN and its China
extension
Since the WiGig/IEEE 802.11ad enables multi-Gbps CH1 CH2 CH3 CH4
throughput, it is expected to spread as a high-speed wireless
alternative to existing wired standards such as HDMI or 57 58 59 60 61 62 63 64 65 66 [GHz]
USB 3.0 as illustrated in Fig. 1(a). Even though the first Fig. 2 Frequency allocations in 60GHz band.
target will be to replace the wired high-speed interface, the
60 GHz wireless is also capable of providing high-speed Table 2 60 GHz regulations among different regions.
multiuser access that can be deployed in dense small cell Output power (Po)
EIRP
networks as in train stations shown in Fig. 1(b). However, and antenna gain (Ga)
since the WiGig/IEEE 802.11ad are based on the CSMA/CA Po ≤ 10 dBm (10 mW)
Japan Not specified
Ga ≤ 47dBi
protocol, the throughput is significantly degraded under
1. Ga < 51 dBi
multiuser environment, necessitating the extension of the EIRP = 82 dBm – 2 (51- Ga)
existing protocol in order to maintain high throughput for Outdoor
US 2. Ga ≥ 51 dBi
multiuser access. EIRP = 82dBm
This section starts with a brief overview of the PHY and Indoor 40 dBm Po ≤ 27 dBm (0.5 W)*
MAC specifications in WiGig/IEEE 802.11ad, and then 43 dBm or 1. Directional ant.
explains future extension activities for the 60 GHz WLAN. 57 dBm (Fixed Po ≤ 27 dBm (0.5 W)
Korea
point-to-point) 2. Omni directional ant.
The section also describes hardware implementations Po ≤ 20 dBm (0.1 W)
including a mobile terminal and an AP prototype.
EU 40 dBm* Not specified
*Specified in spectrum density for a narrow band signal
(b) High-speed multiuser wireless access The most recent global frequency allocation at 60 GHz is
Fig. 1 60 GHz use case examples. illustrated in Fig. 2. Frequency bands around 60 GHz are
available worldwide. Table 2 summarizes 60 GHz
regulations among different regions. As can be seen in the
table, the maximum antenna power allowed in Japan is 10
dBm, which is lower than other regions. It is currently under
discussion for relaxing the regulation from the viewpoint of
global uniformity1. The output power and the antenna gain
vary depending on the process technology and antenna
configurations. Even though GaAs or SiGe processes can
1
The new regulation is expected to be effective in 2015 where the
maximum output power up to 24 dBm (250 mW) will be allowed.
4
cost CMOS solution is preferred for most of consumer Control PHY STF CEF Header Data Subfields
(MCS 0)
applications. In general, the output power in today’s 3.6s 0.7s 4.7s 4.2-297.6s 0.0-45.4s
advanced CMOS technology is limited to much lower than
10 dBm due to transistors’ reliability issue. To extend the /2-BPSK /2-BPSK/QPSK/16QAM
communication distance, the beamforming technology Single Carrier STF CEF Header Data Subfields
(MCS 1 - 12)
which consists of an array antenna with beam steering 1.2s 0.7s 0.6s 0.3-1997.5s 0.0-45.4s
capability is widely used to obtain higher antenna gain, at Fig. 3 Packet structure examples.
the cost of increased power dissipation. A recent publication
[40] employs 16 Tx and 16 Rx array structure, Beacon Interval
demonstrating 4.6 Gbps throughput at 10 m communication BHI DTI
distance while dissipating 960 mW in Rx and 1190 mW in BTI A-BFT ATI CBAP1 SP1 SP2 CBAP2
Tx excluding the power consumption of the baseband signal Time
processing. Fig. 4 Example of access periods within beacon interval.
In WiGig/IEEE 802.11ad, both SC modulation and
OFDM modulation have been adopted, considering various AP
#n
#1
#2
#3
modulation and coding schemes (MCSs) where MCS 0 to
#n
#1
#2
#3
MCS 4 are mandatory. MCS 0 is exclusively used to transmit Responder
control channel messages employing differential BSPK Sector Sweep
Sector Sweep
ACK
Time
Quasi-omni
(DBPSK) modulation with code spreading to ensure better
robustness. The SC modulation supports up to 4.62 Gbps
(/2-16QAM) whereas the OFDM modulation achieves Fig. 5 Example of beamforming training.
6.75 Gbps (64-QAM) maximum PHY throughput rate.
Figure 3 shows the packet structure examples. The The WiGig/IEEE 802.11ad employs bidirectional
packets consist of a short training field (STF) and a channel sequence of beamforming training frame transmission that
estimation field (CEF), followed by a header, data and uses sector sweep and provides the necessary signaling. The
subfields. The subfields can be used for fine beamforming beamforming protocol involves sector-level sweep (SLS),
training. beam refinement protocol (BRP) and beam tracking (BT).
To achieve multi-Gbps throughput, the MAC layer Figure 5 illustrates an example of the SLS, which is the most
requires many modifications from the one in the traditional basic type of Tx beam direction training. During the BTI,
IEEE 802.11. For example, a minimum interval time the initiator sends sector sweep frames as training signals
between transmission packets, called short inter frame space while the responder measures the quality of the received
(SIFS), is shortened from 16 s in the IEEE 802.11a/g to 3 frames using a quasi-omni beam pattern which is the widest
s in the WiGig/IEEE 802.11ad. Many timing parameters of beam width attainable. This can be realized by simply using
random backoff in CSMA/CA are also shortened. On the one antenna element which achieves wider beam width
other hand, the maximum payload size is extended from compared to an array antenna. Then, the responder
2304 octets to 7920 octets. These modifications reduce feedbacks the best sector ID while changing the responder’s
packet overhead and improve transmission efficiency. beam patterns. Similar process can also be applied to align
As illustrated in Fig. 4, channel access occurs during Rx beam direction if needed. The BRP phase after the SLS
beacon intervals and is coordinated using a schedule. The is to enable iterative refinement of its antenna configuration
beacon interval comprises a beacon header interval (BHI) (or antenna weight vector) of both the transmitter and the
and a data transfer interval (DTI). The BHI consists of a receiver. Both SLS and BRP use their own special frames
beacon transmission interval (BTI), and optional association for BF training. In contrast, the BT is implemented by
beamforming training (A-BFT) or announcement adding training fields at the end of a data packet, which is
transmission intervals (ATI). The DTI can include one or used to adjust the antenna configuration during data
more scheduled service periods (SPs) and contention-based transmission.
access periods (CBAPs). In a typical setup, the beacon
interval is set to be around 100 ms [19]. Since the 2.2 Next generation of IEEE802.11ad
propagation characteristic of 60 GHz communications is
highly directional, the Tx and Rx antenna beam pattern need The use of high frequencies comes with both advantages
to be aligned in the right direction to obtain sufficient gain and disadvantages. Large path loss, as well as high
in the BHI. attenuation due to obstacles such as a human body, limits 60
5
GHz applications to those suitable for short-range point-to- of 2.16 GHz per channel, the transceiver suffers from large
point communication. On the other hand, higher frequencies frequency dependent distortion caused by gain variations of
lead to smaller sizes of RF components including antennas, analog circuits as well as multipath environment. To
enabling compact realization of an array structure which minimize performance degradations, the transceiver
offers larger antenna gain with high directivity as mentioned employs sophisticated digital calibration schemes, such as
earlier. These properties unique to 60 GHz help reducing built-in Tx in-band calibration and an Rx frequency domain
interference among terminals, and offering opportunities for equalizer (FDE). These techniques relax the requirement for
realizing high speed point-to-multipoint or multipoint-to- high speed analog circuits, leading to less power
multipoint connections as a future extension. consumption with minimum hardware overhead [41].
Since communication distance at 60 GHz is limited, each The RFIC is fabricated in a 90 nm CMOS process with
AP forms a small cell with 3 to 10 m radius and the multiple 3.1 x 3.75 mm chip size. The RFIC consumes 347 mW in
APs will be used for sufficient area coverage. In such a 60 the Tx mode and 274 mW in the Rx mode. As shown in Fig.
GHz wireless network, the AP cooperative beamforming 7, the RFIC is integrated in the cavity structure antenna
will be highly required to enhance system throughput as well module with 10 x 10 mm small size. Each Tx/Rx antenna is
as to alleviate link disconnection due to shadowing. These printed on the surface of the antenna module which consists
enhancements require efficient beamforming protocol, of four patch elements, providing 6.5 dBi gain with 50
resource management, etc., as it will be explained in more degree beam width. The output power is set to be 2 dBm,
details throughout this paper. which corresponds to 6 dB back-off from the saturated
Standardization for the next generation 60 GHz wireless output power of 8 dBm, providing 8.5 dBm EIRP. The
has already started in IEEE 802.11WG and it is under relatively low output power is chosen to avoid transistor
discussion in the Next Generation 60 GHz Study Group performance degradation due to voltage stress in order to
(NG60SG) 2 . The study group aims to achieve maximum guarantee 10-year lifetime. The measured Rx NF is 7.1 dB.
throughput of 20 Gbps while maintaining backward The antenna module is mounted on the 7x4 cm evaluation
compatibility and coexistence with legacy WiGig/IEEE board with a baseband IC including PHY and MAC layers.
802.11ad, targeting various use cases such as multiuser mass The board is connected to a laptop PC through USB 3.0
data distribution, wireless backhaul, ultra-short-range interface, achieving over 1.7 Gbps maximum MAC
communication, etc. NG60SG also aims to improve the (2.5Gbps PHY) throughput at MCS 9.
spatial reuse of simultaneous nearby transmissions to
increase the aggregated system throughput. It is expected to 3.75mm
LNA QDEM
multiuser MIMO and interference suppression for high LO
SW PLL
throughput multiuser operation. The introduction of the new PA QMOD
2.3 Prototype Hardware of WiGig/IEEE802.11ad WLAN Fig. 6 Block diagram (left) and die photo (right) of RFIC.
2
In May 2015, Task Group 11ay (TGay) has been approved and it took
over the standardization activity from NG60SG to develop a new standard
of IEEE 802.11ay.
6
5bit
area with reduced power consumption, we adopt RF phase a
asinq
Antenna
I
q
shifter (PS) approach where only the low noise amplifier IB
acosq
I 5bit
I TX
(LNA) and the power amplifier (PA) need to be duplicated.
11mm
5bit
The main challenge of this approach is the implementation QB
RX
QB
of an RF PS that operates at 60 GHz band. Figure 8 (a) and 12mm
(b) show the PS circuit and the block diagram of the RFIC (a) (c)
[42]. The RFIC employs RF PSs for 4Tx/4Rx signal paths. Fig. 8 (a) Phase shifter circuits, (b) block diagram of RFIC
The PS utilizes a quadrature hybrid with a digitally and (c) picture of antenna module.
controlled vector combiner. The transformer-based
magnetic coupling hybrid offers a small footprint with 2.0
excellent amplitude/phase accuracy. The vector combiner is 1.8
1.6 MCS 9
Throughput [Gbps]
composed of 5-bit current-controlled RF amplifiers, 1.4 (2.5Gbps PHY)
providing 32 x 32 variable outputs. By flipping the polarity 1.2
of differential IQ signals, it covers one of the four quadrants 1.0
MCS 5
0.8
in the IQ signal space, providing 5° phase resolution over a 0.6
(1.25Gbps PHY)
360° variable range. 0.4
The RFIC is fabricated in a 40 nm CMOS process and it 0.2
is integrated in the miniaturized antenna module shown in 0.0
0 200 400 600
Figure 8 (c). The antenna module is 11 x 12 mm and it Distance [cm]
achieves 7 dBi maximum antenna gain, delivering the total (a) Throughput vs. Communication distance
Tx power of 10 dBm EIRP. Even though the four Tx path
configuration allows to achieve higher output power, the 400 2.0
1.8
output power is set to 3 dBm in the prototype to mitigate 350
Throughput [Gbps]
1.6
thermal constraints. Simple codebook based beamforming 1.4
Distance [cm]
300
offers about 120° beam steering range with 7 steps in the 1.2
250 1.0
azimuth direction. 0.8
200
Figure 9 shows measured MAC throughput over the air. 0.6
0.4
In this measurement, two evaluation boards are fixed to the 150
0.2
moving positioners that can control both communication 100 0
-60 -40 -20 0 20 40 60
distance and their direction. As shown in Fig. 9 (a), the Angle [deg]
measured MAC throughput achieves 1.7 Gbps (MCS 9) up (b) Area coverage (with MCS 9)
to 3 m and 1.0 Gbps (MCS 5) up to 5 m in the maximum Fig. 9 Measured MAC throughput over the air.
antenna gain direction. The area coverage is also measured
by rotating one of the boards in the azimuth direction. Beam
direction is controlled based on the WiGig/IEEE 802.11ad STA 120 degree
BF protocol. As shown in Fig. 9(b), it achieves 1.7 Gbps Antenna
Antenna Module
MAC throughput at more than 2.5 m communication Module
AP 22cm
distance while covering +/- 60 deg angle range.
By using the fabricated antenna modules, a prototype of Top view
STA
the AP is designed. As illustrated in Fig. 10, the combination STA
Exterior Internal unit
of the three antenna modules, covers 360° around the AP,
while supporting up to three concurrent communication Fig. 10 AP structure with three antenna modules (left) and the picture
links with spatial sharing. It may be required to extend of the AP prototype (right).
communication distance to cover larger area by one AP. This
is realized by increasing the number of RF signal paths to
deliver higher output power as well as to enhance antenna
gain, which is a subject of future works.
7
WiGig AP BBU frames, these APs and STAs can measure RSSI from the AP
Optical fronthaul #1 without any interference. After the transmission of sector
sweep frames from AP #1, AP #2 starts to send sector sweep
WiGig AP RRH #1 WiGig AP RRH #2 frames and RSS values from AP #2 is measured by AP #1
(mmw AP #1) (mmw AP #2)
and all STAs. When transmission of sector sweep frames
from all mmw APs is completed, each STA initiates to send
its sector sweep frames according to the order specified in
sector sweep frames sent from mmw APs. Sector sweep
STA #1 STA #3 frames sent from each STA is also utilized to notify the best
STA #2
sector ID for each AP. After the transmission of sector
Fig. 11 Distributed RRH architecture for mmw APs coordination. sweep frames from STA, each AP replies sector sweep
feedback frame to notify the best sector to the corresponding
2.4 Extension to multi-mmw AP coordination by using STA. After transmission of sector sweep and sector sweep
WiGig RRH architecture feedback frames from all STAs, mmw APs send RSS
indicator feedback (RSSI feedback) signal to notify the best
Multiple mmw APs are required to be installed in the sectors and RSS values of all links to the BBU. By using
future mmw WLANs to fully cover a target environment and RSSI feedback signal, BBU calculates signal to interference
overcome shadowing problem. However, installing pulse noise ratio (SINR) for all combinations of best sectors
numerous number of mmw APs imposes several as shown in Fig. 13. Based on the calculated SINRs, the
coordination challenges due to the exhaustive BF training BBU selects AP-STA combinations that satisfy the
with CSMA/CA random access. This subsection presents an minimum required SINR as possible candidates used for the
example of multi-mmw AP coordination for future mmw transmission of RTS, CTS and DATA frames during DTI.
WLAN, which we proposed in [29] [30]. The key idea of Selected combinations are notified to mmw APs by sending
this mmw WLAN is to distribute WiGig AP RRHs in the candidate link information (CLI) signal.
target environment and control them in a coordinated The explained coordinated sector sweep incurs significant
manner via WiGig AP BBU. By employing RRH signaling overhead especially at BHI to avoid interference
architecture, synchronously coordinated BF among multiple between BF trainings. For example in the case of 𝑁 APs
mmw APs can be achieved by simply extending the current and 𝑁 STAs, 𝑁 times period for sector sweeps is needed
MAC protocol shown in Sec. 2.1. From now on, we call compared to the single AP and single STA transmission,
WiGig AP RRH as mmw AP and WiGig AP BBU as BBU while on the other hand at most 𝑁 simultaneous
for simpler notation. Figures 11 and 12 illustrate an example transmissions are available. With rough estimation, the total
of the coordinated mmw WLAN and its MAC protocol, throughput of 𝑁 simultaneous transmission 𝑅𝑁 can be
respectively. In this example, two mmw APs are connected estimated as:
to a BBU through optical fronthaul and three stations
(STAs) coexist in the same coverage area. 𝑅𝑁 = (1 − 𝑁 × 𝛼1 ) × 𝑁 × 𝑅1 , (1)
where 𝑅1 and 𝛼1 are average throughput and overhead
2.4.1 Coordinated Beamforming Training ratio in the case of single AP and single STA transmission,
and it is assumed that 𝑁 × 𝛼1 < 1. From the equation, it is
In traditional MAC protocol defined in IEEE 802.11ad obvious that the scheme is effective when 𝛼1 is small. This
and given in Sect. 2.1, each client STA associates with the implies that, in the proposed protocol, DTI should be long
AP achieving the highest received signal strength (RSS) enough compared to the BHI, however it sacrifices tolerance
value. In contrast, in the MAC protocol shown in Fig. 12, against the mobility. Another coordination technology that
during the BHI, each STA tries to associate with all mmw is also able to reduce the overhead will be discussed in Sect.
APs that have RSS values higher than a specified minimum 3 where location based beamforming is applied by using Wi-
threshold, and the best transmission sectors for all links Fi fingerprint.
between APs and associated STAs are determined. At first,
to initiate the BTI, the BBU sends a trigger frame for starting 2.4.2 Coordinated DATA Frame Transmission
sector sweep to the mmw APs. All mmw APs and STAs
keep silent until receiving the trigger frame. After receiving In DTI, prior to transmission of DATA frames, to reserve
the trigger frame, the representative mmw AP pre- the radio resource and detect the disconnection of the links
determined by the BBU sends sector sweep frames, and all due to human shadowing, all mmw APs and associated
the other APs and STAs measure the RSS values using a STAs exchange RTS/CTS frames. For example, as shown in
quasi-omni beam pattern. In the example shown in Fig. 12, Fig. 12, when three STAs are associated with two mmw APs,
AP #1 is a representative AP and it sends sector sweep the exchanges of six RTS/CTS frames (RTS #1-#6, CTS #1-
frames. Since all other APs and STAs (AP #2 and STA #1- #6) are required. To shorten the duration required for
#3) still keep silent during the transmission of sector sweep exchanging the all RTS/CTS frames, the BBU selects the
8
WiGig AP RRH (mmw AP) replies an ACK frame to the corresponding mmw AP. As
#1 #2 #1 #2 #1 #2 shown in Fig. 12, in the proposed MAC layer protocol, to
reduce the frame transmission error as much as possible,
STAs transmit their ACK frames in a time division manner,
STA #1 #2 #3 STA #1 #2 #3 STA #1 #2 #3 and the time to send ACK frames are specified in API signal.
#1 #2 #1 #2 #1 #2
To increase the MAC efficiency of the proposed protocol,
simultaneous transmission of ACK frames from multiple
STA #1 #2 #3 STA #1 #2 #3 STA #1 #2 #3
STAs can be introduced thanks to the space division
#1 #2 #1 #2 #1 #2 multiple access based on beamforming. When the mmw
AP does not receive the ACK frame, the mmw AP decides
that the link employed for the transmission of DATA frame
STA #1 #2 #3 STA #1 #2 #3 STA #1 #2 #3 is blocked and notifies it to the BBU. Then, the BBU re-
#1 #2 #1 #2 #1 #2
coordinates the allocation of link employed for the re-
transmission of DATA frame. Since mmw AP coordination
enables WLAN to dynamically select the mmw AP
STA #1 #2 #3 STA #1 #2 #3 STA #1 #2 #3
employed for the transmission of DATA frames to the
Fig. 13 Combinations of the links between mmw APs and STAs. associated STA in accordance with the disconnection of the
links and received SINR, it avoids the degradation of
combination of links which minimize the required duration
throughput caused by human shadowing and collision due to
among all possible combinations. When the mmw AP
hidden STA problem.
cannot receive CTS frame from the associated STA, the
mmw AP decides that the link is disconnected and notifies it
3. Wi-Fi/WiGig Coordination for Sub-Cloud
to the BBU by sending a blocking information (BLI) signal.
WLAN
After receiving BLI signal, the BBU determines the link (i.e.
mmw AP) employed for the transmission of DATA frame to
Since the Wi-Fi alliance has integrated the WiGig alliance,
each associated STAs among the remaining combinations of
the future WLAN chipsets must support multi-band such as
the links. Several types of scheduling algorithms can be
2.4, 5, and 60GHz. In IEEE 802.11ad, FST [19] is
introduced to determine the links employed for the
introduced for enabling STAs to transfer the ongoing data
transmission of DATA frames. Determined links are notified
transmission from one band to another band based on the
to the mmw APs and the associated STAs by sending
RSS of the different bands. IEEE 802.11ad also introduces
assigned mmw AP information (API) signal. After the
single MAC address (virtual MAC address) for both Wi-Fi
transmission of API signal, each mmw AP transmits its
and WiGig to facilitate the FST operation. This FST works
DATA frame to the corresponding associated STAs. Finally,
effectively only in the case of single AP where handover is
if the DATA frame is correctly received, the associated STA
occurred only between the homothetic coverage of Wi-Fi
and WiGig. In the case of multiple APs, we should consider coordination in the sub-cloud WLAN.
Another technology is the introduction of C/U splitting
AP Controller Fingerprints DBs
NAV duration
Collect learning info.&
WiFi RSS
WiFi M. Resp.
WiFi M. Req.
WiFi M. Resp.
Distribute DBs to APs via GbE
WiFi M. Req.
fingerprints DB DIFS DIFS DIFS
ON
ON
WiGig best
G bit Ethernet
sector ID DB ・・・ ・・・
Dual-band (5GHz/60GHz)
5 GHz t
BF TXOP
WiGig APs 5 GHz ・・・
DATA
DATA
BRP
DATA
DATA
DATA
DATA
DATA
DATA
BRP
FBK
BRP
ACK
ACK
ACK
ACK
ACK
ACK
ACK
ACK
・・・ ・・・
・・・
Broadcast the information of
A bad beam 60 GHz
The best
the selected best beam via WiFi t
estimated beams
The existing : SBIFS : MBIFS : SIFS
The best beam data links 60GHz 5GHz Move/Green frames: frames from AP/UE
Fig. 15 Dual-band MAC protocol for coordinated WLAN.
Dual-band (5GHz/60GHz) STAs (UEs) [35] over dual-band MAC protocol in the sub-cloud WLAN,
where control frames to be shared with multiple APs are
Fig. 14 Sub-cloud WLAN system architecture. transmitted by Wi-Fi while the high speed data frames are
the handover between APs as well as avoiding interference transmitted by WiGig concurrently from multiple mmw APs.
between the beams from different APs. In this section, we The control frames to be transmitted by Wi-Fi includes the
introduce more comprehensive coordination between Wi-Fi frames needed for Wi-Fi fingerprint measurement, a
and WiGig to realize future high capacity multi-Gbps Network Allocation Vector set (NAVset) frame to avoid
WLANs by using sub-cloud architecture. collision during the beam refinement, and a best beam
identification (BID) frame to inform the selected beam to be
used. Since the coverage of Wi-Fi is much larger than that
3.1 Sub-cloud WLAN using Wi-Fi/WiGig Coordination
of WiGig, the Wi-Fi frames are used to manage location of
UEs, to avoid packet collision between mmw APs, and to
Figure 14 shows system architecture of the sub-cloud inform beam ID information selected autonomously in each
WLAN using Wi-Fi/WiGig coordination [26]. In Fig. 14, mmw AP. Such a tight coordination among Wi-Fi and
multiple dual-band (5 and 60 GHz) APs are connected to an WiGig enables concurrent transmission of multiple data
AP controller (APC) via gigabit Ethernet links. This sub- frames from different mmw APs, and also realizes seamless
cloud WLAN will be installed in a target environment to handover among mmw APs as well as between Wi-Fi and
cover it using multiple dual-band APs. In this architecture, WiGig. It is noted that most of the data traffic is offloaded
the APC works as a central coordinator of the sub-cloud to to WiGig, so that the additional control traffic on Wi-Fi is
assist radio resource management of multiple APs by almost negligible and collisions on control frame hardly
keeping the random access protocol in Wi-Fi/WiGig happen.
networks. Moreover, the APC acts as an interface between
the sub-cloud and cellular networks that will be explained in
3.2 Dual-band MAC Protocol for Coordinated mmw
Sec. 4. To facilitate the coordination process among dual-
Transmissions
band APs, several technologies are used in the proposed
WLAN.
The first technology is an extension of the current FST to Figure 15 shows the dual-band MAC protocol, which we
use Wi-Fi RSS (fingerprint) for location (beam) proposed to coordinate multiple mmw transmissions in
management of the UEs (APs). If more than two APs are random access scenarios [27]. In this protocol, if a data
installed in the target environment, we can roughly identify frame is generated for a specific UE, the random access
the location of a UE by reading Wi-Fi RSS values of process starts using Wi-Fi interface instead of mmw. In the
multiple APs that is called fingerprint in this paper. Since the beginning, one of unused APs senses the 5GHz band using
best AP to be associated and the best beam to be selected are carrier sense routine. If the medium is free, it starts the
location (fingerprint) dependent, the exhaustive search BF backoff counter. If the counter reaches zero, it starts to send
training of multiple APs is not needed anymore if we have a a Wi-Fi measurement request (Wi-Fi M. Req.) frame to
database (DB) to make links between fingerprints and best intended UE. Then, the UE broadcasts a Wi-Fi measurement
beam IDs. For that purpose, offline statistical learning [28] response (Wi-Fi M. Resp.) frame for measuring its current
is introduced in this paper. Moreover, the bad beams from Wi-Fi fingerprint. Based on the current Wi-Fi fingerprint
other APs to make destructive interference to the current and offline fingerprint DBs, the APC associates the UE to an
AP-UE link can also be estimated by using the fingerprint appropriate unused AP. Then, it sends a switch ON frame to
and DB, that bad beams should be eliminated from the BF the UE from the unused AP using its Wi-Fi interface to turn
training in other APs. Thus, efficient coordinated ON the mmw interface if it’s in a sleep mode. The APC also
beamforming among multiple mmw APs without exhaustive estimates a group of best beams for the selected AP-UE link
search BF training can be achieved by Wi-Fi/WiGig and groups of bad beams that produce interference to this
link from other APs. Then, the APC sends this best and bad
10
5 -25
3.3.1.1 Collecting Fingerprints Databases (DBs)
Wi-Fi RSS values in dBm
-30
4
-35
The first step in the offline statistical learning phase is to
Y [m]
3 -40
construct the Wi-Fi and mmw radio maps for target
-45
environments. Constructing the radio maps can be 2
-50
effectively done by collecting the average Wi-Fi RSS
1 -55
readings (fingerprints) and mmw APs best sectors IDs at
-60
arbitrary learning points (LPs) in the target environment. 0
0 2 4 6 8 10 12
Therefore, three databases are constructed as radio maps, i.e.
X [m]
the Wi-Fi fingerprint DB 𝚿, the best sector ID DB 𝚽, and
the offline received power DB 𝐏𝐎𝐅𝐅 , which are defined as: Fig. 16 Wi-Fi RSS fingerprint radio map for AP at X = 10 m, Y = 0.5 m
and Z = 3 m.
𝜓11 ⋯ 𝜓𝐿1 𝜙11 ⋯ 𝜙𝐿1
𝛙=( ⋮ ⋱ ⋮ ),𝚽 = ( ⋮ ⋱ ⋮ ),
𝜓1𝑁 ⋯ 𝜓𝐿𝑁 𝜙1𝑁 ⋯ 𝜙𝐿𝑁
11
6
-20
arbitrary position r, is measured by the APs and collected by
5
the APC. 𝝍𝑟 is defined as:
4
-30 𝝍𝑟 = [𝜓𝑟1 𝜓𝑟2 … . 𝜓𝑟𝑁 ]𝑇 . (4)
-35
Using 𝝍𝑟 , 𝛙, 𝚽 and 𝐏𝐎𝐅𝐅 , the APC can associate the UE
Y [m]
3 -40
-45
to an appropriate unused AP n. After that, the APC estimates
2
-50
a group of best beams for AP n to communicate with the UE
1 -55
at its current position r. This can be done by calculating
-60
Euclidian distance between 𝝍𝑟 and the Wi-Fi fingerprint
0
0 2 4 6 8 10 12 exemplars of each best sector ID 𝑑𝑛∗ . Therefore, a vector of
X [m] Euclidian distance is obtained with a length up to the total
number of best sector IDs. Then, the APC sorts the obtained
Fig. 17 Wi-Fi RSS fingerprint radio map for AP at X = 0.5 m, Y = 0.5 m
and Z=3 m.
vector of Euclidian distance in an ascending order, and it
selects a group of best sector IDs (best beams) 𝑑𝑛∗ (1: 𝑋) up
to 𝑋, where X is the total number of estimated best beams,
6 as:
7
5 𝑑∗ 2
𝑑𝑛∗ (1: 𝑋) = sort (arg min ‖𝝍𝑟 − 𝕴𝑗 𝑛 ‖ )|
Mmw best sector ID
6 ∗
. (5)
𝑑𝑛 1≤𝑗≤C𝑑∗
4 𝑛 1:𝑋
5
After estimating the best beams 𝑑𝑛∗ (1: 𝑋), the APC pre-
Y [m]
3
4 estimates bad beam candidates from other APs m (𝑚 ≠ 𝑛)
2
that may collide with these estimated best beams. In
3
1 evaluating bad beam candidates, we used the following
2 criterion:
0
0 2 4 6 8 10 12 ∗ (𝑥)
𝑑𝑛 𝑑 ∗
X [m] 𝑑𝑏𝑚 (𝑖) = ∀𝑑𝑚,𝑚≠𝑛
∗ (𝑀𝐶𝑆𝑑𝑛∗𝑚(𝑥) < 𝑀𝐶𝑆𝑑𝑛∗ (𝑥) ),
𝑑 ∗ (𝑥) 𝑑∗
The actual mmw coordination transmissions take place in where 𝑝𝑛𝑧𝑛 and 𝑃𝑚𝑧𝑚 are the offline power received at
this real-time phase. During this phase, the online Wi-Fi overlapped LP z from AP n using best beam 𝑑𝑛∗ (𝑥) and from
∗
RSS fingerprint vector 𝝍𝑟 from the target UE, at an AP m using best beam 𝑑𝑚 respectively, 𝜎 2 is the noise
power, and Z is the total number of overlapped LPs. If at
12
least one of the overlapped LPs satisfies (6), the ID number slightly degrade. This comes from the random access in the
∗
of 𝑑𝑚 is registered as a bad beam candidate from AP m un-coordinated exhaustive search BF, which causes a lot of
∗ (𝑥)
𝑑𝑛
to 𝑑𝑛∗ (𝑥)
, i.e. 𝑑𝑏𝑚 (𝑖) . After estimating 𝑑𝑛∗ (1: 𝑋) and packet collisions as previously explained. On the other hand,
𝑑 ∗ (𝑥) using the proposed coordinated mmw transmissions, the
𝑛
𝑑𝑏𝑚 , the APC sends 𝑑𝑛∗ (1: 𝑋) to AP n, and 𝑑𝑛∗ (1: 𝑋) system throughput is highly increased as we increase the
𝑛 𝑑 ∗ (𝑥)
and 𝑑𝑏𝑚 to the AP m. After NAV set using Wi-Fi, AP n number of APs. Using 8 mmw APs, about 6 times increase
conducts BF refinement to find out the best beam 𝑑𝑛∗ (𝑥 ∗ )
among 𝑑∗𝑛 (1: 𝑋), then it broadcasts its ID using the BID
frame using Wi-Fi. By knowing the actual 𝑑𝑛∗ (𝑥 ∗ ), other
∗ (𝑥 ∗ )
𝑑𝑛
APs can find out more accurate bad beams 𝑑𝑏𝑚 (𝑖) using
WiFi/mmw
WiGig APAP8
8
above criterion. If AP m is selected for associating another
𝑑𝑛∗ (𝑥 ∗ ) WiGig AP
WiFi/mmw AP77 WiFi/mmw
WiGig APAP6
6
UE, it will eliminate the estimated bad beams 𝑑𝑏𝑚 from WiFi/mmw
WiGig AP AP55
∗ (1:
its estimated best beams 𝑑𝑚 𝑋) before starting BF WiFi/mmw
WiGig APAP4
4
refinement. Thus, a collision free coordinated BF training WiFi/mmw
WiGig APAP3
3 WiFi/mmw
WiGig APAP2
2
can be achieved. WiFi/mmw
WiGig AP AP11
4m
3.4 Simulation Analysis
In this section, the efficiency of the proposed sub-cloud Fig. 19 Ray tracing simulation area.
WLAN compared to the conventional un-coordinated mmw
WLAN is verified via computer simulations. Figure 19
shows ray tracing simulation area of a target environment, Table 4 Simulation parameters.
in which 8 dual band (Wi-Fi/mmw) APs are deployed on the Parameter Value
ceiling. Simulation parameters are given in Table 4. The Transmission mode SC/FDE
steering antenna model defined in IEEE 802.11ad [19] is Number of mmw APs/UEs 8/24
used as the transmit antenna directivity 𝐺(𝜑, 𝜃) for mmw Tx power of 5 GHz/60 GHz 20dBm/10dBm
APs. In which, the beam gain is designed as 25dBi with half Beamwidth in azimuth and elevation 30° , 30°
power beamwidth of 30° both in azimuth 𝜑 and elevation Traffic model Poisson distribution
𝜃 directions. Using 𝐺(𝜑, 𝜃), the total channel gain from Offered load /Packet size for UE 1Gbps /1500 octet
mmw AP in the mmw band becomes: Max. num. of re-transmission 10
Num. of estimated best beams 6
2𝜋 𝜋
𝑔(𝜏) = ∫0 ∫0 √𝐺 (𝜑, 𝜃) ℎ(𝜑, 𝜃, 𝜏)sin𝜃d𝜃d𝜑, (9) Num. of antenna sectors 36
Num. of LPs 90
where ℎ(𝜑, 𝜃, 𝜏) is the channel response calculated by the Beacon interval time 1sec
ray tracing simulation without BF gain. Beam gain 25 dBi
In the simulation, the following two metrics are evaluated.
The first one is total system throughput, which is defined as
the sum of throughputs for all successfully delivered packets
from all mmw APs. The second one is the average packet Un-coordinated mmw WLAN
delay, which is defined as average time period from the 12 Coordinated mmw WLAN
Average total throughput [Gbps]
400
350
300
250
200
Fig. 22 3GPP LTE/WLAN internetworking architecture.
150
1 2 3 4 5 6 7 8
Number of APs by anybody, and therefore we call these WLANs as third-
Fig. 21 Average packet delay. party radio access networks (3RANs). In order to efficiently
operate 3RANs in LTE areas, the LTE uses ANDSF to
in total system throughput is achieved by the proposed partially control the offloading process through the WLANs.
coordinated WLAN over the un-coordinated one. Likewise, ANDSF is used to support users to efficiently discover and
the proposed coordinated mmw WLAN greatly enhances the select 3RANs autonomously by giving a list of nearby
average packet delay. Also, using 8 APs, the average packet WLANs over the LTE link as a very primitive version of
delay of the un-coordinated WLAN is 2.5 times longer than C/U splitting [37]. ePDG works as a gateway for users to
that of the proposed coordinated WLAN. This packet delay establish end-to-end secure connections to packet data
reduction comes from not only avoiding packet collisions network (PDN), e.g. Internet, through evolved packet core
but also eliminating the exhaustive search BF mechanism (EPC) [44]-[46]. Although this HetNet architecture has
using statistical learning. several advantages, such as simplicity and good cost
performance, it lacks tight coordination over WLANs.
4. LTE/WiGig Internetworking for 5G cellular Especially in the case of mmw WLANs, since the coverage
Networks of WLANs are very small, it is difficult to coordinate them
Although interworking between high capacity WLANs via loosely coupled architecture due to the latency of the
and LTE networks is a key enabler of future 5G cellular signalling. Energy efficient WLAN discovery protocols also
networks, there are several challenges to efficiently cannot be achieved because users themselves scan through
internetwork them. These challenges include WLAN the list of nearby WLANs to finally decide to which ones
discovery, coordination over WLANs, and financial issues they associate. Therefore, optimal WLAN selection cannot
in terms of capital expenditures (CAPEX) and operating be done due to the lack of centralized control. As well,
expense (OPEX). The HetNet architecture and C/U splitting schemes based on data pre-fetching and delayed offloading
including the function of ANDSF are proposed to facilitate cannot be handled [47] [48].
the LTE and WLAN internetworking [32]-[37]. By using
these techniques, the wide coverage LTE is used to perform 4.2 Tightly Coupled LTE/WLAN Internetworking
the signaling and a combination of LTE and WLAN
opportunities is used to deliver the data. In this section, we
Figure 23 depicts the tightly coupled LTE/mmw WLAN
introduce three different architectures for LTE/WLAN
architecture proposed by the authors in [38]. In this
internetworking namely the loosely coupled in the current
architecture, one evolved node B (eNB), namely LTE base
3GPP standard [36], the tightly coupled, and the hybrid
station (BS), and multiple mmw WLANs are directly
coupled LTE/WLAN internetworking proposed by the
connected and are in master-slave relationships in a unified
authors in [38] and [39].
evolved universal terrestrial RAN (E-UTRAN) architecture.
Therefore, concrete C/U splitting can be easily operated and
4.1 Loosely Coupled 3GPP LTE/WLAN Internetworking managed using this architecture. Hence, radio resources and
WLANs context information can be uniformly managed and
The main objective of the current 3GPP architecture [36], conveyed by the eNB. Control signals are transmitted solely
shown in Fig. 22, is to actively facilitate user data offloading by the eNB, while user data could be opportunistically
from LTE to WLANs, by which traffic congestion in LTE served by either the eNB or WLANs. In order to realize
could be relaxed. Due to this simple objective, the seamless LTE/mmw handover via C/U splitting, the
architecture is very simple and mainly depends on three architecture employed two LTE/WLAN protocol adaptors in
independent parts: WLANs, ANDSF and evolved packet mmw WLAN and UE. The function of the adaptors is to
data gateway (ePDG), as shown in Fig. 22. WLANs are mutually translate and encapsulate IP-conformant
independently and directly connected to the Internet. This LTE/WLAN signalling. Mmw WLANs have one half of the
means that WLANs can be installed anywhere and owned protocol adaptors so that the eNB and mmw WLANs are
14
Fig. 23 Tightly LTE/WLAN internetworking architecture. Fig. 24 Hybrid LTE/WLAN internetworking architecture.
able to communicate using X2 reference point defined by multi-layer HetNet is monitored and controlled by E-
3GPP [49]. Another half of the protocol adaptors are ANDSF in each EPC and mmw WLAN APC. Many WLAN,
installed on UEs, by which UEs are able to handle mmw MNO and user context information can be conveyed using
controlling signals conveyed by the eNB and user data E-ANDSF for concrete control over the suggested HetNet
transmitted by mmw WLANs using upper-layer processors architecture. As a result, WLAN discovery and WLAN
for mmw WLAN and LTE, respectively. optimal selection can be achieved in a tree type centralized
The strength of this architecture is that the mmw WLANs manner. Furthermore, data prefetching, cache control, user
are able to be members of uniformed cloud RANs (C-RANs) trajectory expectation, delayed offloading and cell
[32] [33]. Uniformed C-RANs may provide uniformed user broadcasting can be easily offered.
experience regardless of radio access technologies (RATs) to To regulate 3RAN usages among the MNOs, the authors
be used, e.g. seamless handover from LTE to WLANs and proposed to lease and share resources at 3RANs in traffic
vice versa. One of the potential challenges, on the other hand, basis. Thus, MNOs pay for the amount of actually served
is a costing issue for RAN operation. This is because all of traffic or reserved traffic capacity and upper-stream traffic
the BSs, the WLANs and internetworking facilities have to flows are regulated as contracted. This regulation model
be constructed and maintained solely by a single mobile enables MNOs to moderately satisfy high quality services
network operator (MNO) – this challenge becomes and high user experience, compared to the architecture given
significant especially if the number of WLANs increases. in Sect. 4.1, with much lower cost than that in Sect. 4.2,
Furthermore, standardization of control signalling between thanks to IRX and E-ANDSF. More reliable network
the eNB and mmw WLANs is known to be a future work. credibility and protocols for inter-RAN cooperation in terms
of authentication, authorization, and accounting could be
4.3 Hybrid Coupled LTE/WLAN Internetworking important topics for further investigations.
works will contribute to the realization of 5G systems, [13] Australian Communications and Media Authority (ACMA), “Radio
especially in 2020 Tokyo Olympic. communications (Low Interference Potential Devices) Class License
2000 as amended,” 19 Jun. 2013.
Acknowledgments
[14] Korea Communications Commission (KCC) rules, “Korea
This work was partly supported by “The research and Communications Commission (KCC) rules,” 2011.
development project for expansion of radio spectrum [15] Japanese Ministry of Internal Affairs and Communications (MIC),
resources” of The Ministry of Internal Affairs and Radio Regulatory Council, “Radio Law,” 2011.
Communications, Japan. Also, Part of this work has been [16] Chinese Ministry of Industry and Information Technology
done under a project named “Millimeter-Wave Evolution for Departments (MIIT), “Notification on issues of micro power (short
Backhaul and Access (MiWEBA)” under international
range) radio technology applications in 60 GHz spectrum,” Radio
cooperation program of ICT-2013 EU-Japan supported by
FP7 in EU and MIC in Japan. Administration Article, 2006.
[17] Industry Canada, “Spectrum Utilization Policy for License Exempt
References Wireless Devices in the Bands 46.7-46.9 GHz, 57-64 GHz, and 76-77
GHz,” Jan. 2001.
[1] CISCO Whitepaper, “CISCO Visual Networking Index: Global [18] Wireless Gigabit Alliance, Inc., “WiGig MAC and PHY Specification,
Mobile Data Traffic Forecast Update, 2014-2019,” Feb 2015. Version 1.2,” Mar. 2013.
[2] Lu Lu, G.Y. Li, A.L. Swindlehurst, A. Ashikhmin and Rui Zhang, [19] IEEE 802.11ad standard: “Enhancements for Very High Throughput
“An Overview of Massive MIMO: Benefits and Challenges,” IEEE J. in the 60 GHz Band,” Dec. 2012.
of Sel. Topics in Sig. Proces., Vol. 8, Issue 5, pp. 742 – 758, April [20] http://www.wi-fi.org
2014. [21] J. Wang et al., “Beam Codebook Based Beamforming Protocol for
[3] R. Q. Hu, Y. Qian, S. Kota and G. Giambence, “Hetnets - a New Multi-Gbps Millimeter-wave WPAN Systems,” IEEE J. Sel. Areas
Paradigm for Increasing Cellular Capacity and Coverage [Guest Commun., vol. 27, no. 8, pp. 1390-1399, Oct. 2009.
Editorial],” IEEE Wireless Commun., vol. 18, no. 3, pp.8-9, Jun. 2011. [22] K. Hosoya et al., “Multiple sector ID capture (MIDC): A novel
[4] K. Sakaguchi et al “Millimeter-wave Evolution for 5G Cellular beamforming technique for 60 GHz band multi-Gbps WLAN/PAN
Networks,” IEICE Trans. Commun, vol. E98-B, no. 3, pp.338-402, systems,” IEEE Trans. Antennas Propagat., vol .63, no. 1, pp. 81-96,
Mar. 2015. Jan. 2015.
[5] D. Liu, U. Pfeiffer, J. Grzyb and B. Gaucher, “Advanced Millimeter- [23] http://www.qca.qualcomm.com/thewire/newspresslist/qualcomm-
wave Technologies: Antennas, Packaging and Circuits,” Wiley, 2009. shows-wigig-prototype-tablets-snapdragon-810-processors/
[6] A. Sadri, “Evolution of mmWave Technology from WiGig to cellular [24] Ehab M. Mohamed, K. Sakaguchi and S. Sampei, “Delayed
and Backhaul Systems,” Brooklyn 5G Summit, Apr. 2014 Offloading using Cloud Cooperated Millimetre Wave Gates,” in Proc.
[7] A. Maltsev et al “MmWave Smallcells is a Key Technology for IEEE PIMRC, Sept. 2014, pp. 1852-1856.
Future 5G Wireless Communication Systems,” European Conference [25] Ehab M. Mohamed, K. Sakaguchi and S. Sampei, “Delayed
on Networks and Communications (EuCNC'2014), Jun. 2014. Offloading Zone Associations using Cloud Cooperated
[8] T. S. Rappaport et al., “Broadband Millimeter-Wave Propagation Heterogeneous Networks,” in Proc. IEEE WCNC workshops, Mar.
Measurements and Models Using Adaptive-Beam Antennas for 2015, pp. 322-327.
Outdoor Urban Cellular Communications,” IEEE Trans. Antennas [26] H. Kusano, E. M. Mohamed, K.Sakaguchi, S.Miyamoto, and S.
Propagat., vol.61, No.4, pp. 1850-1859, Apr. 2013. Sampei, “Cooperative Resource Control of Millimeter Wave Wireless
[9] ERC Recommendation 70-03, "Relating to use of Short Range Access via Wi-Fi Coordinator”, TR. of IEICE, SR 2013-117,
Devices," 9 Oct. 2012. Mar.2014.
[10] ETSI EN 302 567, “60 GHz Multiple-Gigabit WAS/RLAN Systems; [27] E. M. Mohamed, H. Kusano, K.Sakaguchi, S.Miyamoto, and S.
Harmonized EN covering the essential requirements of article 3.2 of Sampei, “Wi-Fi Assisted Dual-band MAC Protocol for Coordinated
the R&TTE directive,” Nov. 2011. Beamforming among WiGig APs in Indoor Environment”, TR. of
[11] ECC Recommendation (09)01, “Use of the 57 - 64 GHz Frequency IEICE, SR 2014-138 , Mar. 2015.
Band for Point-to-Point Fixed Wireless Systems”. [28] E. M. Mohamed, K.Sakaguchi, and S. Sampei, “Millimeter Wave
[12] FCC ET Docket No. 07-113, Report & Order, “Revision of Part 15 of Beamforming Based on Wi-Fi Fingerprinting in Indoor Environment,”
the Commission’s Rules Regarding Operation in the 57-64 GHz Submitted for ICC workshops, Jun. 2015.
Band”. [29] S. Miyazaki, S. Miyamoto, K. Sakaguchi, Seiichi Sampei, “A
Proposal of Access Point Coordinated Media Access Control Protocol
16
for Millimeter Wave Wireless LAN Systems,” IEICE Technical [46] 3GPP TS 33.402 V12.5.0: “3GPP System Architecture Evolution
Report, RCS2014-334, pp.195-200, Mar. 2015. (SAE); Security aspects of non-3GPP accesses”.
[30] S. Miyazaki, S. Miyamoto, K. Sakaguchi, Seiichi Sampei, “A Study [47] K. Lee, J. Lee, Y. Yi, I. Rhee and S. Chong, “Mobile Data Offloading:
on Beamforming Training Interval for Supporting Low-Speed Mobile How Much Can Wi-Fi Deliver?”, Networking, IEEE/ACM
Station in Access Point Coordinated Millimeter-Wave WLAN Transactions on, vol.21, no.2, pp.536-550, Apr.2013.
Systems,” Proc. of IEICE Gen. Conf.’15, B-5-153, Mar. 2015. [48] V. A. Siris and M. Anagnostopoulou, “Performance and Energy
[31] J. Qiao, X. Shen, J. W. Mark, and Y. He, “MAC-layer Concurrent Efficiency of Mobile Data Offloading with Mobility Prediction and
Beamforming Protocol for Indoor Millimeter Wave Networks,” IEEE Prefetching”, in Proc. IEEE WoWMoM, pp.1-6, 4-7 Jun. 2013.
Trans. On Vehicular Technology, vol. 64, no. 1, pp. 327-338, Jan. [49] 3GPP TS 36.420, “E-UTRAN, X2 General Aspects and Principles,”
2015. v12.0.0, Jun. 2014.
[32] China Mobile Research Institute, “C-RAN - Road Towards Green
Radio Access Network,” Wireless World Research Forum, Meeting Kei Sakaguchi received the B.E. degree in
electrical and computer engineering from
27 (WWRF27), Dusseldorf, 2011.
Nagoya Institute of Technology, Japan in
[33] Next Generation Mobile Networks (NGMN) Alliance, “Suggestions 1996, and the M.E. degree in information
on Potential Solutions to C-RAN by NGMN Alliance,” Jan. 2013. processing from Tokyo Institute of
Technology, Japan in 1998, and the Ph.D.
[34] Next Generation Mobile Networks (NGMN) Alliance, “Smallcell degree in electrical and electronic
Backhaul Requirements,” Jun. 2012 engineering from Tokyo Institute of
[35] H. Ishii, Y. Kishiyama and H. Takahashi, “A novel Technology in 2006. From 2000 to 2007, he
was an Assistant Professor at Tokyo Institute
Architecture for LTE-B C-plane / U-plane Split and Phantom Cell of Technology. Since 2007, he has been an
Concept,” in Proc. Globecom workshops, 2012. Associate Professor at the same university.
Since 2012, he has also joined in Osaka
[36] 3GPP TS 23.234 V12.0.0: “3GPP system to Wireless Local Area
University as an Associate Professor. He
Network (WLAN) interworking; System description”. received the Young Engineering Awards from
[37] 3GPP TS 24.312 V12.7.0: “Access Network Discovery and Selection IEICE and IEEE AP-S Japan Chapter in 2001
and 2002 respectively, the Outstanding Paper
Function (ANDSF) Management Object (MO)”.
Awards from SDR Forum and IEICE in 2004
[38] H. Peng, T. Yamamoto and Y. Suegara, “LTE/WiGig RAN-level and 2005, respectively, the Tutorial Paper
Award from IEICE Communication Society
Interworking Architecture for 5G Millimeter-wave Heterogeneous
in 2006, and the Best Paper Awards from
Networks,” submitted to IEICE Special Section on 5G Radio Access IEICE Communication Society in 2012,
Networks – Part II. 2013, and 2015. He is currently playing roles
of the General Chair in IEEE WDN-CN2015,
[39] M. Mizukami, K. Sakaguchi and S. Sampei, “Novel Network the Honorary General Chair in IEEE
Architecture and Business Model for Consolidated LTE/WLAN CSCN2015, and the TPC Chair in IEEE
Heterogeneous Network with Tight Cooperation,” IEICE Tech. Rep., RFID-TA2015. His current research interests
are 5G cellular networks, sensor networks,
vol. 114, no. 490, RCS2014-321, pp. 117-122, Mar. 2015. and wireless energy transmission. He is a
[40] M. Boers et al “A 16TX/16RX 60GHz 802.11ad Chipset with Single member of IEEE.
Coaxial Interface and Polarization Diversity,” in IEEE ISSCC Dig.
Ehab Mahmoud Mohamed received the
Tech. Papers, Feb. 2014, pp.344-345.
B.E. degree in electrical engineering from
[41] T. Tsukizawa et al, “A fully integrated 60 GHz CMOS Transceiver South Valley University, Egypt, in 2001, and
Chipset Based on WiGig/IEEE802.11ad with Built-in Self- the M.E. degree in computer science from
South Valley University, Egypt in 2006, and
Calibration for Mobile Applications,” in IEEE ISSCC Dig. Tech.
the Ph.D. degree in Information Science and
Papers, Feb. 2013, pp.230-231. electrical engineering from Kyushu
[42] K. Takinami et al, “A 60GHz wireless transceiver employing hybrid University, Japan in 2012. From 2003 to
2008, he was an Assistant Lecturer at South
analog/digital beamforming with interference suppression for Valley University, Egypt. Since 2012, he has
multiuser Gigabit/s radio access”, Symposium on VLSI Circuits, Dig. been an Assistant Professor at Aswan
Tech. Papers, pp.306-307, June 2015. University, Egypt. Since 2013, he has also
joined in Osaka University, Japan as a
[43] B.J. Frey and D. Dueck, “Clustering by passing messages between Specially Appointed Researcher. He is a
data points,” Science, vol. 315, no. 1, pp. 972-976, Feb. 2007. technical committee member in many
international conferences and a reviewer in
[44] 3GPP TS 23.402 V13.0.0: “Architecture enhancements for non-3GPP
many international conferences and
accesses”. transactions. His current research interests are
[45] 3GPP TS 24.302 V13.0.0: “Access to the 3GPP Evolved Packet Core 5G networks, cognitive radio networks,
millimeter wave transmissions and MIMO
(EPC) via non-3GPP access networks; Stage 3”.
17
Naganori Shirakata received B.E and M.E. Propagation (EuCAP 2011) in 2011.
degrees in electronics engineering from
Kyoto Institute of Technology, Japan in 1993
and 1995 respectively. He joined Panasonic
Corporation, (former Matsushita Electric
Industrial Co., Ltd.), Osaka in 1995, where he
worked on baseband signal processing of
OFDM and MIMO for WLANs. From 2008
to 2010, He worked on ultra low power radio
and WBAN systems. From 2010 to 2013, he
worked on the development of the PHY
system and circuit design for millimeter wave
transceiver. He currently leads the
development of the system design for
millimeter wave access point and network.
He is a member of IEEE.