Received January 5, 2020, accepted January 15, 2020, date of publication January 20, 2020, date of current version

January 29, 2020.

Digital Object Identifier 10.1109/ACCESS.2020.2967747

Small Cells Enabled by Crowdsourced Radio Units

Mounted on Parked Vehicles for Smart City
YU NAKAYAMA 1,2 , (Member, IEEE), TAKAYUKI NISHIO 3 , (Member, IEEE),
DAISUKE HISANO 4 , (Member, IEEE), AND KAZUKI MARUTA 2,5 , (Member, IEEE)
1 Institute
of Engineering, Tokyo University of Agriculture and Technology, Tokyo 184-8588, Japan
2 NPO neko 9 Laboratories, Tokyo 140-0001, Japan
3 Graduate School of Informatics, Kyoto University, Kyoto 606-8501, Japan
4 Graduate School of Engineering, Osaka University, Osaka 565-0871, Japan
5 Graduate School of Engineering, Chiba University, Chiba 263-8522, Japan

Corresponding author: Yu Nakayama (yu.nakayama@ieee.org)

This work was supported in part by the JST ACT-I under Grant JPMJPR18UL, and in part by the Mazda Foundation, Japan.

ABSTRACT A large number of small cells are densely deployed in the era of 5G and beyond 5G mobile
networks to achieve high rate and low latency data transmission. Small cells are composed with fixed ground
radio units (RUs) in the conventional networks, and thus the utilization rate of them is drastically deteriorated
due to the spatio-temporal patterns of mobile traffic demand. To address this problem, adaptive network
architectures have been investigated to reconfigure networks on demand using small cells mounted on
moving objects such as vehicles and drones. However, it has been difficult to establish stable communication
link to ground nodes in high-speed environments with these schemes. Therefore, this paper proposes a
concept of small cells composed with crowdsourced RUs mounted on parked vehicles. They are activated
based on the traffic demand in surrounding area by road side units (RSUs) assuming a smart city. The
proposed idea is based on the correlation between the distribution of people and the occupancy rate of parking
lots. It can efficiently improve user throughput with relatively small number of RUs. The feasibility of the
proposed scheme is validated through numerical analysis, computer simulations, and experimental results.

INDEX TERMS Mobile communication, wireless communication, communication system, vehicles.

I. INTRODUCTION are densely deployed to compose small cells. Densifying

Along with the spread of various mobile services includ-
ing cellular communication and Internet of things (IoT),
the amount of mobile traffic has been increasing rapidly.
In the era of 5G and beyond 5G mobile networks, a large
number of small cells are densely deployed for widespread
mobile devices to achieve high rate and low latency data
transmission. A popular and prevailing network architecture
for efficiently deploying many small cells is the centralized
radio access network (C-RAN) architecture [1]. The func-
tions of a mobile base station (BS) are implemented sep-
arately in three components in the C-RAN architecture as
shown in Fig. 1. They are called a central unit (CU), a dis-
tributed unit (DU), and a radio unit (RU). An RU provides
the antenna, analog-to-digital converter (ADC) and digital-
to-analog converter (DAC) functions, a DU does other phys-
ical (PHY) layer and medium control access (MAC) layer
functions, and a CU does higher layer functions. Many RUs
The associate editor coordinating the review of this manuscript and
approving it for publication was Antonino Orsino . daytime than nighttime. The future increase in the mobile data
small cells is the most straightforward approach to enhance
the system capacity, because the performance of transmit
power control (TPC) and radio resource management (RRM)
control approaches is limited by the possible power emission
for distant user equipments (UEs). RUs communicate with a
CU pool in a central office via DUs. A link between a CU
and a DU is called midhaul which requires large bandwidth.
A link between a DU and an RU is called fronthaul, where
strict latency requirements are defined, e.g. ≤ 100 µs [2].
For the efficient deployment of a large number of fronthaul
links, wireless links have been studied in addition to con-
ventional optical links [3]. To ensure the security of 5G
mobile networks, network coding (NC)-enabled small cells
are proposed for intrusion detection and prevention [4].
Not only the increase in the amount, there are signifi-
cant spatio-temporal fluctuations in the mobile traffic. These
demand fluctuations result from the patterns of human mobil-
ity such as commuter movement [5], [6]. For example, there
are far more mobile phone users in business districts during

This work is licensed under a Creative Commons Attribution 4.0 License. For more information, see http://creativecommons.org/licenses/by/4.0/
17986 VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 1. C-RAN architecture.
rates will intensify these demand fluctuations. The utilization
ratio of network equpiments is drastically deteriorated by
demand fluctuations with conventional small cells where the
system capacity is fixed. Thus, it has become an important
research topic to efficiently deploy a large number of small
cells dealing with demand fluctuations.
An adaptive network architecture has been investigated to
address this issue [7], [8]. In this architecture, moving nodes
relocate following the changes in the demand distribution
to continuously provide connectivity for mobile users with
small number of deployed nodes. As deployment scenarios of
this architecture, the concept of optically backhauled moving
network [9], [10] and vehicle-mounted small cells were pro-
posed [11], [12]. In particular, the concept vehicle-mounted
cells is a promising solution for the efficient deployment of
mobile networks, because it can utilize the power of citizens.
The on-board RUs, which are also called as crowdsourced
RUs (CRUs), are activated on demand to compose small cells
and relay traffic between DUs and UEs. The drivers receive
monetary or non-monetary incentive from network operators
as reward for forwarding mobile traffic. The activation states
of CRUs are determined by drivers or edge servers based
on traffic demand, power consumption, and incentive. The
vehicle-to-pedestrian (V2P) connectivity between on-board
RUs and mobile users in moving environment with high
frequency band was evaluated in [13]. However, in these
schemes the utilization of moving vehicles was only con-
sidered. As regards private vehicles spend much more time
in parking lots than on roads in reality, it is reasonable to
focus on parked vehicles, which have been out of scope in
the previous works.
Therefore, in this paper we propose a concept of on demand
activation of small cells with CRUs mounted on parked
vehicles for smart city. The proposed idea is based on the
correlation between mobile traffic demand and occupancy
rate of neighboring parking lots. Assuming a smart city
VOLUME 8, 2020
where road side units (RSUs) are widely deployed, edge
servers grasp the distribution, states, and battery charge rate
of CRUs vie RSUs. Edge servers optimally control the states
of CRUs based on the collected information. With the pro-
posed scheme, stable wireless fronthaul links can be eas-
ily established with static channels by focusing on parked
vehicles in stable environments. It contributes to efficient
deployment of mobile networks thanks to its high flexibility.
The basic idea was proposed in [14], [15]. This paper provides
more comprehensive description on the proposed scheme,
simulation results, and experimental results.
The rest of the paper is organized as follows. Section II
describes related works and the contribution of this work
in detail. The proposed scheme is explained in section III.
Section IV clarifies the applicable condition of the pro-
posed scheme via numerical analysis. The feasibility of the
proposed idea is evaluated through computer simulations
and experiments. The simulation results are described in
section V. Section VI introduces the experimental results. The
conclusion is provided in section VII.
II. RELATED WORK
Moving small cells has been a hot research topic for efficient
deployment of small cells. A scenario where small cells are
deployed on top of public transportation vehicles such as
buses and taxis was investigated in [16]. The traffic generated
by mobile users inside the vehicles and in neighboring areas is
carried by the vehicle mounted cells. They confirmed that the
moving cells were efficient in crowded areas where they can
provide connectivity for stationary traffic hotspot. The per-
formance of moving cells mounted on public buses was eval-
uated in [17]. In this scenario, a moving cell aggregates user
traffic and communicate with a ground node using a wireless
backhaul link. Due to the limited resources such as wireless
resource blocks and energy, resource allocation has been a
research issue for moving small cells. A joint allocation prob-
lem was proposed in [18] to enhance the quality-of-service
(QoS) in cellular networks considering power consumption.
To support mobile backhaul connectivity for vehicle mounted
cells, a multi-Gbps vehicular communication system was
proposed in [19]. It achieves multi-Gbps data rate transmis-
sion for train and car applications by efficiently allocating
resources in moving environments that create time-varying
interference using millimeter-wave (mmWave) frequency
band. It can be pointed out that these schemes have focused on
providing connectivity for the passengers in moving vehicles.
The concept of crowdsourced RUs was proposed for
dynamically enhancing mobile networks to provide con-
nectivity for mobile users in neighboring areas [11], [12].
CRUs are mounted on small vehicles such as private cars
and taxis, which does not compose traffic hotspot itself. The
on-board RUs are activated on demand to compose small
cells. The vehicle-to-pedestrian (V2P) connectivity between
mobile users and CRUs in moving environment with high fre-
quency band was evaluated in [13]. It was confirmed via com-
puter simulations that multiple moving cells can be utilized
17987
 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
for a coordinated multipoint (CoMP) transmission/reception
to expand coverage. The goal of this concept is to enhance
mobile connectivity for users inside and outside vehicles on
demand with crowdsourcing, i.e. utilization of the power of
citizens. As the reward for forwarding mobile traffic, the vehi-
cle drivers receive monetary or non-monetary incentive from
mobile network operators. It is assumed that the of each CRU
is activated/deactivated by vehicle drivers or edge servers
in accordance with traffic demand, power consumption, and
incentive. However, these schemes focused on vehicles in
moving environment; it is rather difficult for moving CRUs
to establish stable wireless fronthaul links with ground DUs.
Therefore, this paper proposes a concept of on demand
activation of small cells with CRUs mounted on parked
vehicles [14], [15]. The proposed idea is based on a sim-
ple fact that it is easier to establish stable link with parked
vehicles than moving vehicles. It leverages the correlation
between traffic demand and the number of parked vehicles.
The proposed scheme assumes the existence of edge servers
that control the activation states of CRUs because there are
various types of drivers [20]. The contribution of this work
is to propose a promising approach for efficiently deploy
numerous small cells of mobile networks in 5G and beyond
5G era. The feasibility of the proposed scheme is clarified
with analytical, simulation, and experimental results.
III. PROPOSED NETWORK ARCHITECTURE
A. CONCEPT
The concept of the proposed small cells composed by CRUs
mounted on parked vehicles is depicted in Fig. 2. A part
of conventional static ground RUs are replaced by CRUs
installed on parked vehicles. They are activated to compose
small cells in addition to conventional ground RUs to enhance
the network capacity. They are connected with DUs via
FIGURE 2. Small cell with CRUs mounted on parked vehicles.
17988
FIGURE 3. Mobile demand and occupancy rate.
wireless fronthaul links. The power of a CRU is supplied by
the car battery. It is assumed that the wireless fronthaul links
basically employ CPRI which is a widely used interface in the
C-RAN architecture. In addition to this, the legacy decode-
and-forward (DF) relaying based on LTE/WiMAX/WiFi stan-
dards can also be applicable. The major limitation for relaying
is the strict latency requirement of fronthaul. Main point is
that the CRUs can be activated or de-activated in accordance
with the mobile traffic demand.
The utilization scenario for the proposed idea is shown
in Fig. 3. The proposed concept utilizes the general corre-
lation between the distribution of mobile users and the occu-
pancy rate of parking lots in the same area. A typical example
of this trend emerges in commercial stores such as restaurants
and coffee shops. During daytime, mobile users park their
private vehicles and spend time in the building of these stores.
During nighttime, on the contrary, few or no mobile users
are in the building, i.e. the store is closed, and consequently
the parking lots are vacant. The correlation between traffic
demand and parked vehicles is also applicable to residential
areas where users drive home from work at night. Considering
the correlation described here, if mobile network operators
install ground RUs in each area to satisfy the peak demand,
the capacities of deployed RUs are underutilized for most
of time.
The proposed scheme assumes a smart city where RSUs
are widely deployed. We also assume that it employs an edge
computing architecture where edge servers grasp the distribu-
tion, states, and battery charge rate of CRUs vie neighboring
RSUs. Edge servers optimally activate/deactivate CRUs via
RSUs based on the collected information. The advantages of
the proposed scheme are high flexibility and low cost. These
are enabled by the feature of the proposed network archi-
tecture that the distribution of CRUs follows the distribution
of mobile users. As a consequence, the proposed idea can
drastically reduce the number of ground RUs while dealing
with the peak traffic rate in each area.
VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
B. PRELIMINARY RESULT
Here we present a preliminary examination to ensure the
effectiveness of the proposed concept. Throughput perfor-
mance is compared with moving environments through a
simple link level simulation. Physical (PHY) layer parameters
assume LTE based orthogonal frequency division multiplex-
ing (OFDM) transmission with 20 MHz bandwidth in which
the number of FFT points is 2048 and 1200 subcarriers are
available. Subcarrier spacing is 15 kHz and a short cyclic
prefix (CP) is inserted; the symbol duration is 71.4 µs. Adap-
tive modulation and coding (AMC) scheme is also employed
to maximize the throughput performance in accordance with
signal-to-noise power ratio (SNR). Modulation orders are
QPSK, 16QAM, 64QAM and 256QAM. Turbo code rates
are 1/2, 2/3, 3/4 and 5/6. Suppose one time slot transmis-
sion which contains 1 pilot symbol and 6 data symbols,
throughput performance was evaluated with regard to the
mobility. Time-varying multipath fading channel is based
on traditional Jakes’ model associated with tapped delay
line type A (TDL-A) with 300 ns root mean square (RMS)
spread [21]. The angular distribution of scattering is assumed
to be uniform from all azimuth directions. Vehicle’s mobility
is represented by the max Doppler frequency fD , e.g. 40 km/h
(25.9 mph) speed corresponds to fD = 130 Hz at 3.5 GHz
carrier frequency.
Fig. 4 shows link level throughput performance with
Doppler frequencies of 0 and 130 Hz, respectively. As shown
in the figure, the achievable throughput is reduced by 59.9%
from the static case. This result indicates that it is easier to
achieve stable and satisfactory communication using vehicle-
mounted CRUs in static environments than those in high
mobility environments. Therefore, the proposed concept is
quite reasonable to better support traffic hotspot in a flexible
manner.
C. SYSTEM ARCHITECTURE
Fig. 5 depicts the system architecture of the proposed scheme.
Ground RUs and DUs are deployed in the same way as
FIGURE 4. Link level throughput performance.
VOLUME 8, 2020
FIGURE 5. System architecture.
conventional C-RAN. In addition to them, CRUs are mounted
on small vehicles such as private cars, buses, and taxis. An
edge server is placed to establish control channels with widely
deployed RSUs. RSUs are expected to be essential in a smart
city for providing intelligent transportation systems (ITS) and
vehicle-to-vehicle (V2V) communication. After a vehicle is
parked in a parking lot, the on-board CRU checks-in the
neighboring RSU; the RSU checks the real-time distribu-
tion of CRUs and their states in its coverage area. Based
on the collected information received from RSUs, the edge
server computes the activation states of CRUs. According
to the computed results, CRUs mounted on parked vehicles
are either activated or deactivated via RSUs. If a CRU is
activated, it establishes a wireless mobile fronthaul link with
a neighboring DU. An activated CRU composes a small cell
overlaid on small cells composed by conventional ground
RUs to provide mobile connectivity. If the engine of the
vehicle is started, the CRU function is automatically dis-
abled and the CRU checks-out the RSU. It is assumed that
the car owners receive monetary or non-monetary incentive
from a mobile network operator for providing network con-
nectivity. Note that CRUs are controlled by edge servers
to reduce the influence of user behavior; CRUs are usu-
ally on standby for remote activation even when they are
deactivated.
D. BATTERY CHARGE
It is assumed that the power of a CRU is supplied by the car
battery on which it is mounted. Fig. 6 shows a conceptual
diagram of the charging rate of car battery. The blue line
(CRU1) represents an example charging rate for gasoline-
powered vehicles and hybrid vehicles. The green line (CRU2)
is an example charging rate for electric vehicles.
In the case of CRU1, the car battery is charged as the car
drives, and the battery charge stops when the car parks. This
represents general functions of a car battery and a generator;
17989
 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 6. Conceptual diagram of battery charge rate.
the generator generates electrical power and stores it in the
battery. Then, after the activation of the CRU function,
the charged power is used. When the charging rate becomes
less than the predefined threshold, the CRU function is deac-
tivated to save energy. In this case, the proposed approach
can contribute to the better use of energy. This is because the
surplus power generated by the generator can be utilized by
the CRU function.
In the case of electric vehicles which is shown as the green
line (CRU2) in Fig. 6, the charged power is used for driving
and the car battery is charged at charging stations. When the
CRU function is activated, the power consumption slightly
increases. This is because the power consumption of the CRU
function is relatively small compared with the large battery
capacity of electric vehicles.
E. ACTIVATION ALGORITHM
We introduce the proposed CRU activation algorithm in this
section. The proposed algorithm is executed by an edge server
for each coverage area of RSUs in the smart city. Based
on the current information on the traffic demand and CRU
distribution, CRUs are selectively activated.
1) DEMAND OBSERVATION
First, we define a block as a planning unit for the proposed
algorithm. The concept of a block is depicted in Fig. 5. A
block consists of a small area such as a store with parking
lots and residential buildings with shared parking lots. The
size of a block is determined by the cell size of a CRU; the
size of a block can increase in accordance with the increase
in the cell size. Let i denote the identifier for a block.
The mobile traffic demand in the ith block is defined as
di (t). Let pi denote the capacity of the parking lots in the ith
block. The occupancy rate of parking lots in the ith block is
denoted as oi (t). Note that the mobile traffic demand and
occupancy rate are expressed as functions of time, because
they vary with time. Based on the assumption that there is a
correlation between di (t) and oi (t) [22], we formulate
di (t) = wi (t)pi oi (t),
17990
where wi (t) is the correlation parameter. The value of wi (t)
is determined by the characteristics of the block. pi is a pre-
defined value which is determined by the physical capacity
of parking lots. oi (t) can be easily measured in real-time
with existing solutions using cameras or sensors. Equation (1)
represents that the amount of mobile traffic increase in accor-
dance with the number of parked vehicles. That is, there will
be more people, i.e. mobile users, as the parked vehicles
increase, which is pointed in previous work [22].
2) CHECK-IN
When a vehicle is parked at parking lots in the ith block,
the CRU mounted on the vehicle communicates with the RSU
that covers the block, which is referred to as the ith RSU. This
control channel is established with a wireless connection such
as macro cell communication. This procedure is defined as
check-in in this paper.
Let c denote the identifier for a CRU in the ith block. Let
Ci denote the set of candidate CRUs in the ith block, i.e. the
current activatable CRUs in this area. The battery charging
rate of the cth CRU is denoted as bc . Let Bh denote the
minimum charging threshold for activating a CRU, which
is set considering the expected activation time in the block.
After the check-in of the cth CRU, it reports bc to the ith RSU.
If bc satisfies
bc > Bh , (2)
The cth CRU is added to Ci . That is, a CRU with sufficient
battery level is treated as an activatable CRU. Note that too
high Bh results in shortage of activatable CRUs, while too
low Bh causes frequent deactivation and activation of CRUs
and inefficient use of energy. If a parked vehicle is activated
and starts to move, the corresponding CRU checks-out; it is
excluded from Ci .
The RSU periodically send Ci to the edge server. Conse-
quently, the edge the current information on the distribution
of available CRUs in each block.
3) ACTIVATION
The capacity of mobile traffic provided by a CRU is defined
as s. Let ni (t) denote the required number of active CRUs in
the ith block. ni is simply formulated as
 
di (t)
ni (t) = . (3)
s
The edge server selects ni (t) CRUs from Ci based on bc .
The CRUs in Ci is sorted in descending order of bc , and
they are selected in this order. The computed result is sent
to the RSU. The RSU activate the selected CRUs via the
control channels. As a consequence, CRUs with more battery
charging rate are activated to provide stable connectivity.
The activated CRUs periodically send the current battery
level bc to the RSU. Let Bl denote a threshold for deactivating
the CRU function. When the battery charging rate becomes
low and satisfies
(1) bc ≤ Bl , (4)
VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City

The cth CRU is excluded from Ci . The threshold Bl is deter-
mined as regards the battery utilization. If the energy is shared
with other functions of the vehicle, Bl is set to a large value to
ensure the energy for such functions. On the contrary, if the
energy is only used by the CRU function, Bl can be set as 0 to
fully utilize it. Whenever Ci is updated, the edge server selects
another CRU from Ci to fulfill the condition (3).
and hard shutdown ones. The latter assumes users who ter-
minate the standby mode of CRUs for some reason such as
energy-saving. Thus, available CRUs are mounted on γ of
all the parked vehicles and they are included in Ci . The size
of Ci can be formulated as
|Ci | = γ pi oi (t). (8)

With (8), (7) is described as

F. COMPLEXITY  
This paper assumes a smart city where RSUs are widely wi (t)pi oi (t)
γ pi oi (t) ≥ . (9)
deployed and vehicles continuously communicate with these s
RSUs. This type of communication is usually called vehicle Therefore, the requirement for the number of CRUs is
to infrastructure (V2I). V2I communication is expected to always satisfied in the ith block if the parameters satisfy
be essential for autonomous driving and related technologies
in the near future. Based on this reasonable assumption, wi (t)pi oi (t)
γ pi oi (t) ≥ + 1, (10)
a control channel between a CRU and an RSU requires small s
bandwidth and is not time-sensitive. This feature results from which is derived from (9) and and a ≤ dae < a + 1.
the simple communication sequence that consists of check- From (10), the requirement is rewritten as
in, reporting battery charging rate, and activate/deactivate.
wi (t) 1
The proposed activation algorithm itself is also very simple ≤γ − . (11)
and its computational complexity is O(n). Thus, the proposed s pi oi (t)
algorithm can be easily executed in real-time by an edge Fig. 7 shows the applicable region of the proposed
server. Also, if computational resource is available at an RSU, approach as formulated in (11). It is applicable when the
the proposed algorithm can be executed by an RSU without traffic demand is not too large for the number and capacity
communicating with an edge server. of CRUs in the area. The number of CRUs increases with
either the attach ratio of available CRUs (γ ), the parking lots
IV. NUMERICAL ANALYSIS capacity (pi ), or the occupancy rate of parking lots (oi (t)).
The proposed scheme can contribute to dense and flexible Since wi (t) varies with time, it is efficient to ensure large
deployment of small cells unlike the conventional fixed man- s compared with the variation of wi (t) to stably provide
ner. The benefit of the proposed idea is cost reduction through sufficient connectivity. The proposed approach can stably
reduction in the number of deployed RUs. Meanwhile, it is provide spectral resource to traffic hotspot as the number of
difficult for all of private vehicles to mount the CRU function CRUs in parking increases.
in reality. Thus, here we provide a theoretical analysis for the
required attach ratio of CRUs using the parameters defined in B. CRU ATTACH RATIO
section III such as mobile traffic demand. Here the applicable condition for the CRU attach ratio γ
is described. First, it is assumed to simplify the problem
A. APPLICABLE CONDITION that vehicles move around a city where the total capacity of
As formulated in (3), sufficient number of available CRUs is parking lots in the city is equal to the total number of vehicles.
required in each block to satisfy mobile traffic demand with Another assumption is that the power consumption of a CRU
the proposed approach. The requirement for the number of function is low compared to the capacity of car battery, and
available CRUs in the ith block is formulated as thus Bh and Bl are set to negligibly small values. Let Ncru
|Ci | ≥ ni (t). (5) denote the total number of available CRUs in the city. It is
formulated as
This inequality simply implies that the size of Ci must be Nhs
equal to or greater than the required number of CRUs. Also,
X
Ncru = γ pi , (12)
ni (t) can be described using (1) and (3) as i
 
wi (t)pi oi (t) where Nhs denote the number of parking lots, i.e. hotspots in
ni (t) = . (6)
s the city.
Thus, (5) is equivalent to The condition that γ needs to satisfy can be derived
  from (1) and (11) as
wi (t)pi oi (t)
|Ci | ≥ . (7) 1

di (t)

s γ ≥ +1 . (13)
pi oi (t) s
The attach ratio of available CRU on vehicles is defined
as γ . If a CRU is attached on a vehicle, it can be available The mobile demand di and the occupation rate oi are
or unavailable. Unavailable CRUs include low-battery charge expressed as functions of time, because they vary with time.

VOLUME 8, 2020 17991

 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 7. Applicable condition of the proposed scheme.
Thus, the minimum value of γ is described as

1

di (t)

γMin = Max +1 . (14)
pi oi (t) s ground RUs are doubled from Case 1.
As stated above, movable CRUs can be deployed in addi-
tion to conventional fixed ground RUs to efficiently enhance
network connectivity. The number of ground RUs in the ith
block is defined as gi . For simplicity, the cell size and capacity
of mobile traffic of a ground RU are assumed to be the same
with a CRU; the capacity is s. If there are gi ground RUs
installed in the area, the mobile traffic demand that needs to
be supported by CRUs is described as
di0 (t) = di (t) − sgi .
Thus, with the existence of ground RUs, (14) is relaxed with
di0 (t) without loss of generality. It is efficient to minimize
the total number of deployed RUs using ground RUs and
proposed CRUs based on the demand distribution. Therefore,
the minimum value and condition for γ can be formulated as
follows.
X X
Min γ pi + gi (16)
i
i  s = 10, and thus the number of active CRUs was calculated as
1 di (t)
s.t. γ ≥ − gi + 1 ∀i, t
pi oi (t) s in total, the required number of CRUs without ground RUs
V. COMPUTER SIMULATION
The major use case of the proposed scheme is efficiently
dealing with traffic hotspots on demand. The goal of the
mathematical model formulated above is to clarify the appli-
cable condition for the proposed approach. Following com-
puter simulation then validates the proposed system model
by substituting the possible parameter set such as the parking
capacity, the occupancy rate, and the correlation coefficient.
As a result, it is implied that throughput performance can be
improved by additional CRUs while the required number is
fewer than densifying fixed small cells.
A. SIMULATION CONDITION
The condition for computer simulation is explained in the fol-
lowing. We considered following three scenarios for deploy-
ing small cells. For each case, achievable user throughput was
measured.
17992
Case 1: Conventional deployment; Fixed ground RUs are
deployed.
Case 2: Dense deployment; The number of deployed fixed
Case 3: Proposed solution; CRUs mounted on parked vehi-
cles are activated in addition to the ground RUs in
Case 1.
The study area in computer simulation consisted of 1 km2
square region. For the basic condition in Case 1, 200 ground
RUs were uniformly distributed. Also, 200 UEs were uni-
formly distributed in the study area as shown in Fig. 8. In
addition to these UEs, Nhs traffic hotspots were heteroge-
(15) neously originated, which were assumed to be coffee shops
or restaurants next to parking lots. The number of hotspots
which was denoted as Nhs ranged from 5 to 50.
For the proposed scheme, the parking capacity was set as
pi = 25, and the occupancy rate was set as oi = 0.4. It was
assumed that average four people were expected to be in a
vehicle, and thus the correlation parameter was set as wi = 4.
The traffic demand in the ith (i = 1, . . . , Nhs ) block was
derived from (1) as di = 40. The specification of a CRU and
a ground RU was assumed to be the same. Their capacity was
(17) ni = 4 with (3). Since Nhs ni CRUs were additionally activated
was 313 from (12). Considering that the number of deployed
ground RUs was 200, the number of CRUs which were
actually activated was around 100. The above detailed system
parameters give an intuitive understanding that coincides with
the numerical analysis. The attach ratio γ was calculated as
0.16 with (16) and (17).
As regards wireless transmission, each UE was associated
to an RU such that the highest received signal strength is
obtained. When the the vth RU serves Nue,v UEs, the spectral
resource is assumed to be shared among them. Suppose full
buffer traffic, all active RUs transmit signals to designated
UEs in a synchronous manner. Evaluation metric is defined
as the downlink throughput observed at the uth UE;
 
ηB |huu|2
0u = log2 1 + , (18)
v∈Sk , |huv | +σn /ρ
2 2 2
P
Nue,v
v6 =u
VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
TABLE 1. Simulation parameters.
where huv denotes the channel coefficient from the vth RU
to the uth UE. Sk ∈ N contains the set of indices of RUs
to which the kth channel is assigned. B, ρ 2 and σn2 denote
the transmission bandwidth, the (C)RU transmission power
and the UE noise variance, respectively. η = 0.85 represents
the overhead loss derived from the occupied bandwidth ratio
(=0.90) and short cyclic prefix (=0.93) in consideration of
3GPP LTE specification [23]. Channel model considered
long-term fluctuation including shadowing effect. Path loss
model is based on 3D urban micro cell (3D-UMi) scenar-
ios [24]. Standard deviation of shadowing is set to 4 dB. Here
we examine the effect of the multi-channel use. Available
channels are randomly assigned to ground RUs. As for the
CRU, different channels are equally assigned to each acti-
vated vehicle. Detailed parameters related to the wireless
function and propagation model are is based on [24] and
are listed in Table 1. RU antenna height is assumed to be
1.5 m which is the rooftop of the vehicles. UEs are assumed
to locate up to the second floor; their height is determined
among 1.2, 4.2, and 7.2 m at random.
A number of trial is conducted with different RUs/UEs
deployment and statistic throughput characteristics are eval-
uated.
B. SIMULATION RESULTS
The simulation results are described in the following. Fig. 9
shows the cumulative distribution functions (CDFs) of user
throughput. Compared with Case 1, it can surely improve
the system capacity to densify small cells (Case 2). The
throughput improvement at the cell-edge region, i.e. the 5th
percentile baseline, is 38.7% from Case 1. A major key point
of small cell network is that the number of inactive cells
increases in accordance with the densification. That is, UEs
are associated with their nearest RUs, and unused RUs do
VOLUME 8, 2020
FIGURE 8. Example of network layout.
not need to emit radio waves. The received power of the
desired signal can be enlarged whereas that of interference
is reduced [25], [26]. Resultant SINR can be raised and thus
overall system throughput is improved.
The proposed scheme (Case 3) outperforms other cases
by activating small cells near traffic hotspots. The 5th per-
centile performance can be improved by 119.8% and 58.5%
compared with the Cases 1 and 2, respectively. In the sim-
ulative evaluation, the number of ground RUs, which is
denoted as Ngru , for each case was 200, 400, and 200,
respectively. In the Case 3, Ncru = 100 CRUs were addi-
tionally deployed, and thus the total number of RUs for
each case was 200, 400, and 300. This implies that the
proposed scheme can reduce overall RUs by 25% com-
pared with the Case 2. In summary, the throughput per-
formance can be improved by additionally activating CRUs
while the number of them is saved fewer than densifying
fixed ground RUs. If there are additional channels available,
the throughput can be further enhanced. As shown in Fig. 9,
doubling the number of channels can raise the 5th per-
centile characteristics by 72.8%. Detailed characteristics are
discussed below.
Fig. 10 shows the 5th percentile throughput performance
with the number of available channels. User throughput
performance of the proposed approach can be enhanced as
increasing the number of channels while that of the fixed
small cell deployment (Cases 1 and 2) saturates up to dou-
bled channels. It implies that simply densifying ground RUs
cannot well support traffic hotspot even with more number of
channels. CRUs on parked vehicles are adjoiningly located
nearby the traffic hotspots. In this case, although CRUs can
accommodate hotpot users, co-channel interference among
them is inevitable. Channel segregation among CRUs can
effectively mitigate its impact and thus throughput perfor-
mance can be improved. Meanwhile, this study employed
a quite simple channel and UE assignment that considers
only the desired signal strength. More sophisticated radio
17993
 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 9. CDFs of user throughput.
FIGURE 10. 5th percentile user throughput versus number of channels
(Nhs = 25).
resource management strategies are expected to maximize the
achievable throughput performance.
Fig. 11 shows the 5th percentile throughput performance
with the number of hotspots. The figure also plots the results
with different number of channels. Here we observed up to
50 hotspots (Nhs ni = 200) where the proposed concept is ben-
eficial in terms of total number of RUs. It is obvious that the
throughput performances tend to be degraded with increase in
Nhs since the radio resource is shared among UEs. Effect of
the channel segregation is also consistent irrespective to Nhs ;
preparing more number of channels can contribute to enhanc-
ing the system capacity especially for the proposed approach.
It should be noted that the proposed approach with the single
channel shows comparable throughput performance with the
densified fixed ground RUs with multiple channel usage.
Localizing small cell deployment adjacent to traffic hotspot
is quite effective. This observation supports the argument that
our proposal is also spectrally efficient.
17994
FIGURE 11. 5th percentile user throughput versus number of hotspots.
FIGURE 12. Experiment condition.
C. DISCUSSION
There have been many research efforts on the C-RAN archi-
tecture for 5G and beyond. Among them, a major concept for
efficient deployment of small cell has been a moving small
cell. In the previous works such as [16], [17], the performance
of moving small cells was investigated. In particular, they
evaluated the performance in the scenario where moving
small cells were mounted on the top of public transporta-
tion vehicles including taxis and buses. The goal of these
previously presented vehicle mounted cells has been to pro-
vide stable network connectivity for passengers in moving
vehicles. The contribution of this paper is to provide con-
nectivity for mobile users outside vehicles. Since mobile
traffic demand fluctuates in time and generates hotspots,
dynamic small cell deployment is effective for dealing with
such fluctuating demand. The proposed idea is a deployment
scenario for the adaptive network architecture introduced in
[7]. Also, the latency in a wireless fronthaul link which is
the major constraint for mobile fronthaul was investigated
in [12]. The point is that the latency has been confirmed in
the previous work. The improvement in user throughput was
confirmed through computer simulation in this paper. Also,
VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 13. Scenario 1) hotspot condition.
TABLE 2. Equipment specifications.
we confirmed the efficiency of the proposed approach by
comparing dense small cell deployment.
VI. PROOF OF CONCEPT: EXPERIMENT RESULTS
A. EXPERIMENT SETUP
The feasibility of the proposed concept was proven via
outdoor experiment. IEEE802.11ac/ad wireless LAN equip-
ments were used in the experiment; a DU was simu-
lated with IgniteNet’s MetroLinq 60-10G-360, a CRU was
simulated with MetroLinq 60-LW, and UEs were ASUS’s
RogPhone. The equipment specifications are summarized
in Table. 2.
Fig. 12 shows the experiment condition. In the experiment,
two CRUs were mounted on parked vehicles. They were
linked with a neighboring DU with IEEE802.11ad based
60 GHz band millimeter-wave links. The PHY rate was
automatically controlled within the range 4.62 Gbps (max-
imum) and 385 Mbps (minimum). A server was connected
with the DU via 10 GbE. To simulate a conventional ground
RU, the DU also worked as an RU. IEEE802.11ac with
5.6 GHz channel, which is allowed outdoor use, was applied
to radio access links between CRUs and UEs. Two kind of
bandwidths were examined; 80 MHz channel bonding with
866.7 Gbps maximum rate and dual 40 MHz channel bonding
with 400 Mbps maximum rate. Channels were automatically
selected on W56 band. The server sent full buffer UDP or
TCP traffic to UEs through iPerf, and the throughput was
measured at each UE. The experiments were conducted
in two scenarios; 1) Hotspot scenario, and 2) Distributed
scenario.
VOLUME 8, 2020
FIGURE 14. Scenario 1) throughput performance.
B. RESULTS
1) SCENARIO 1
The conditions and results of scenario 1 are depicted in
Figs. 13 and 14, respectively. A hotspot such as a coffee
shop was assumed in this scenario. In the conventional archi-
tecture in Fig. 13b, the fixed DU (i.e. ground RU) directly
serves co-located CRUs whereas the transmission distance
is relatively long; around 40 meters away. The proposed
architecture activates a CRU mounted on a parked vehicle
nearby the hotspot as shown in Fig. 13a. This CRU can pro-
vide stable communication link with the neighboring hotspot
UEs. Resultant throughput performance is shown in Fig. 14.
Both UDP and TCP total throughputs were improved by
120.4% and 80.8% compared with the conventional architec-
ture, respectively. Although the access link is provided via
wireless relay with, 60 GHz band link has larger capacity
than the 5.6 GHz link and thus its bottleneck effect was
negligibly small. Fig. 14 also presents detailed throughput
values per UE. It can be confirmed that equal user throughput
is provided with the UDP case because transmission dis-
tance for all UEs are almost the same. Meanwhile, the TCP
case shows unbalanced throughput due to its congestion
control function, which is not the outcome of the network
architecture.
17995
 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
FIGURE 15. Scenario 2) distributed condition.
FIGURE 16. Scenario 2) throughput performance.
2) SCENARIO 2
The conditions and results of scenario 2 are shown in
Figs. 15 and 16, respectively. This scenario assumes dis-
tributed UEs where six UEs distribute within a coverage
area of a small cell. Conventional fixed ground RU covers
these UEs in the same way with scenario 1 (Fig. 15c). In
this case, user throughput is dependent on the distance as
shown in Fig. 16a. Two CRUs mounted on vehicles were
activated with the proposed approach. In this scenario as
depicted in Figs. 15a and 15b, each UE was connected with
a neighboring RU including a ground RU and the acti-
vated CRUs. In this approach, traffic demand of UEs is
17996
distributed to RUs with which each UE can obtain sufficient
signal strength.
The first trial utilized 80 MHz single channel for all access
links as shown in Fig. 15a. UDP and TCP total through-
puts were improved by 51.9% and 102.4%, respectively.
In the second trial, we separated 80 MHz bandwidth into
two 40 MHz for each CRU as shown in Fig. 15b and its
channel utilization is depicted in Fig. 15d. Despite dividing
the available bandwidth, overall throughput performances
were further improved by 129.1% for UDP and 127.9%
for TCP. 80MHz channel is chosen from 100–112ch or
116–128ch according to the IEEE802.11ac standard whereas
each 40 MHz channel is set to be automatically determined
from 100–112ch and 116–128ch, respectively. Its spectrum
utilization example is marked as red in Fig. 15d. Therefore
co-existence between one of the separated channels for CRUs
and 80 MHz channel for ground RU was avoided. It can
improve MAC efficiency of distributed coordination func-
tion (DCF) to avoid co-channel interference with ground
RUs. The above field experiment verified the effectiveness
of the proposed concept.
VII. CONCLUSION
Prominent spatio-temporal fluctuation in mobile traffic
demand foreseen in the era of 5G and beyond would dete-
riorate the efficiency of conventional mobile networks. In
order to address this problem, we proposed a novel concept of
dynamic C-RAN architecture that crowdsourced radio units
VOLUME 8, 2020
Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
mounted on parked vehicles support traffic hotspots nearby
them. Road side units control neighboring crowdsourced
radio units based on the distribution and battery charge rate
of them. The correlation between the demand and occupancy
rate of parking lots is utilized in the proposed concept; the
applicable region was numerically analyzed. System level
simulation clarified that our proposal is also advantageous in
terms of the user throughput performance as well as reducing
the number of overall small cells. Furthermore, its feasibility
has been proven via experiment; localizing small cell access
points nearby the traffic hotspot can enhance the through-
put performance when high capacity back/fronthaul link is
available. The proposed concept can be one of the promising
solutions to realize a smart city.
ACKNOWLEDGMENT
This article was presented in part at the IEEE 90th Vehic-
ular Technology Conference (VTC-Fall), Honolulu, Hawaii,
USA, September 2019.
REFERENCES
[1] A. Pizzinat, P. Chanclou, F. Saliou, and T. Diallo, ‘‘Things you should
know about fronthaul,’’ J. Lightw. Technol., vol. 33, no. 5, pp. 1077–1083,
Mar. 1, 2015.
[2] Study on New Radio Access Technology: Radio Access Architecture and
Interfaces, document 3GPP TR 38.801, 2017.
[3] D. Hisano, Y. Nakayama, K. Maruta, and A. Maruta, ‘‘Deployment design
of functional split base station in fixed and wireless multihop fronthaul,’’
in Proc. IEEE Global Commun. Conf. (GLOBECOM), Dec. 2018.
[4] R. Parsamehr, A. Esfahani, G. Mantas, A. Radwan, S. Mumtaz,
J. Rodriguez, and J.-F. Martinez-Ortega, ‘‘A novel intrusion detection and
prevention scheme for network coding-enabled mobile small cells,’’ IEEE
Trans. Comput. Soc. Syst., vol. 6, no. 6, pp. 1467–1477, Dec. 2019.
[5] T. Louail, M. Lenormand, O. G. C. Ros, M. Picornell, R. Herranz,
E. Frias-Martinez, J. J. Ramasco, and M. Barthelemy, ‘‘From mobile
phone data to the spatial structure of cities,’’ Sci. Rep., vol. 4, Jun. 2014,
Art. no. 5276.
[6] M. Lenormand, M. Picornell, O. G. Cantú-Ros, A. Tugores, T. Louail,
R. Herranz, M. Barthelemy, E. Frías-Martínez, and J. J. Ramasco, ‘‘Cross-
checking different sources of mobility information,’’ PLoS ONE, vol. 9,
no. 8, Aug. 2014, Art. no. e105184.
[7] Y. Nakayama, T. Tsutsumi, K. Maruta, and K. Sezaki, ‘‘ABSORB:
Autonomous base station with optical reflex backhaul to adapt to fluctu-
ating demand,’’ in Proc. IEEE Int. Conf. Comput. Commun. (INFOCOM),
May 2017.
[8] Y. Nakayama, R. Yasunaga, D. Hisano, and K. Maruta, ‘‘Adaptive network
architecture with moving nodes towards beyond 5G era,’’ in Proc. IEEE
Int. Conf. Commun. (ICC), May 2019.
[9] Y. Nakayama, K. Maruta, T. Tsutsumi, and K. Sezaki, ‘‘Optically back-
hauled moving network for local trains,’’ in Proc. 4th ACM Workshop Hot
Topics Wireless (HotWireless), 2017, pp. 31–35.
[10] Y. Nakayama, K. Maruta, T. Tsutsumi, and K. Sezaki, ‘‘Optically back-
hauled moving network for local trains: Architecture and scheduling,’’
IEEE Access, vol. 6, pp. 31023–31036, 2018.
[11] K. Honda, R. Yasunaga, Y. Nakayama, K. Maruta, and T. Tsutsumi,
‘‘Dynamic mobile network architecture organized by drivers decision
making,’’ in Proc. IEEE 29th Annu. Int. Symp. Pers., Indoor Mobile Radio
Commun. (PIMRC), Sep. 2018.
[12] Y. Nakayama, K. Honda, D. Hisano, and K. Maruta, ‘‘Adaptive C-RAN
architecture for smart city using crowdsourced radio units,’’ in Proc. IEEE
Int. Conf. Pervasive Comput. Commun. Workshops (PerCom Workshops),
Mar. 2019.
[13] K. Maruta, Y. Nakayama, K. Honda, D. Hisano, and C.-J. Ahn, ‘‘V2P con-
nectivity on higher frequency band and CoMP based coverage expansion,’’
in Proc. IEEE 29th Annu. Int. Symp. Pers., Indoor Mobile Radio Commun.
(PIMRC), Sep. 2018.
VOLUME 8, 2020
[14] Y. Nakayama and K. Maruta, ‘‘Adaptive C-RAN architecture for smart city
with crowdsourced radio units mounted on parked vehicles,’’ in Proc. IEEE
90th Veh. Technol. Conf. (VTC-Fall), Sep. 2019, pp. 1–5.
[15] Y. Nakayama, D. Hisano, T. Nishio, and K. Maruta, ‘‘Experimental results
on crowdsourced radio units mounted on parked vehicles,’’ in Proc. IEEE
90th Veh. Technol. Conf. (VTC-Fall), Sep. 2019, pp. 1–2.
[16] A. Jaziri, R. Nasri, and T. Chahed, ‘‘Offloading traffic hotspots using
moving small cells,’’ in Proc. IEEE Int. Conf. Commun. (ICC), May 2016,
pp. 1–6.
[17] M. F. Feteiha, M. H. Qutqut, and H. S. Hassanein, ‘‘Outage probabil-
ity analysis of mobile small cells over LTE—A networks,’’ in Proc.
Int. Wireless Commun. Mobile Comput. Conf. (IWCMC), Aug. 2014,
pp. 1045–1050.
[18] S. Jangsher and V. O. Li, ‘‘Resource allocation in moving small cell
network,’’ IEEE Trans. Wireless Commun., vol. 15, no. 7, pp. 4559–4570,
Mar. 2016.
[19] G. Noh, J. Kim, H. Chung, and I. Kim, ‘‘Realizing multi-Gbps vehicular
communication: Design, implementation, and validation,’’ IEEE Access,
vol. 7, pp. 19435–19446, 2019.
[20] S. Sarwar, S. Zia, Z. U. Qayyum, M. Iqbal, M. Safyan, S. Mumtaz, R.
García-Castro, and K. Kostromitin, ‘‘Context aware ontology-based hybrid
intelligent framework for vehicle driver categorization,’’ Trans. Emerg.
Telecommun. Technol., Aug. 2019, Art. no. e3729.
[21] Study on Channel Model for Frequencies from 0.5 to 100 GHz (Release
16), 3GPP Standard TR 38.901 V16.0.0, 3GPP, Oct. 2019.
[22] F. Mohammadnia, M. Fiore, and M. A. Marsan, ‘‘Adaptive densification
of mobile networks: Exploring correlations in vehicular and telecom traf-
fic,’’ in Proc. 17th Annu. Medit. Ad Hoc Netw. Workshop (Med-Hoc-Net),
Jun. 2018, pp. 1–8.
[23] Physical Layer Aspect for Evolved Universal Terrestrial Radio Access
(UTRA) (Release 7), 3GPP Standard 3GPP TR 25.814 V7.1.0, Sep. 2006.
[24] Study on 3D Channel Model for LTE (Release 12), 3GPP Standard 3GPP
TR 36.873 V12.7.0, Dec. 2017.
[25] M. Kamel, W. Hamouda, and A. Youssef, ‘‘Ultra-dense networks:
A survey,’’ IEEE Commun. Surveys Tuts., vol. 18, no. 4, pp. 2522–2545,
4th Quart., 2016.
[26] S. W. H. Shah, A. N. Mian, S. Mumtaz, and J. Crowcroft, ‘‘System capacity
analysis for ultra-dense multi-tier future cellular networks,’’ IEEE Access,
vol. 7, pp. 50503–50512, 2019.
YU NAKAYAMA (Member, IEEE) received the
B.Agr. degree in agriculture, the M.Env. degree
in environmental studies, and the Ph.D. degree in
information and communication engineering from
The University of Tokyo, Tokyo, Japan, in 2006,
2008, and 2018, respectively. From 2008 to 2018,
he was with NTT Access Network Service Sys-
tems Laboratories. He is currently an Associate
Professor with the Institute of Engineering, Tokyo
University of Agriculture and Technology. He is
also the President of NPO neko 9 Laboratories, which is a nonprofit organi-
zation in Tokyo. His research interests include mobile computing, network
architecture, the IoT, and ultra low latency. He is a member of the IEICE and
IPSJ.
TAKAYUKI NISHIO (Member, IEEE) received the
B.E. degree in electrical and electronic engineer-
ing, and the master’s and Ph.D. degrees in infor-
matics from Kyoto University, in 2010, 2012, and
2013, respectively. From 2016 to 2017, he was a
Visiting Researcher with the Wireless Information
Network Laboratory (WINLAB), Rutgers Univer-
sity, USA. He is currently an Assistant Professor
in communications and computer engineering with
the Graduate School of Informatics, Kyoto Univer-
sity. His current research interests include machine learning-based network
control, machine learning in wireless networks, and heterogeneous resource
management.
17997
 Y. Nakayama et al.: Small Cells Enabled by Crowdsourced Radio Units Mounted on Parked Vehicles for Smart City
DAISUKE HISANO (Member, IEEE) received the
B.E., M.E., and Ph.D. degrees in electrical, elec-
tronic, and information engineering from Osaka
University, Osaka, Japan, in 2012, 2014, and 2018,
respectively. In 2014, he joined NTT Access Net-
work Service Systems Laboratories, Yokosuka,
Japan. Since October 2018, he has been an Assis-
tant Professor with Osaka University. His research
interests include optical-wireless converged net-
works, optical communication, and all-optical sig-
nal processing. He is a member of the Institute of Electronics, Information,
and Communication Engineers (IEICE) of Japan.
17998
KAZUKI MARUTA (Member, IEEE) received the
B.E., M.E., and Ph.D. degrees in engineering from
Kyushu University, Japan, in 2006, 2008, and
2016, respectively. From 2008 to 2017, he was
with NTT Access Network Service Systems Labo-
ratories and also engaged in the research and devel-
opment of interference compensation techniques
for future wireless communication systems. He is
currently an Assistant Professor with the Graduate
School of Engineering, Chiba University. He is
also the Vice-President of NPO neko 9 Laboratories, Tokyo. His research
interests include MIMO, adaptive array signal processing, channel estima-
tion, medium access control, and moving networks. He is a member of the
IEICE. He received the IEICE Young Researcher Award, in 2012, the IEICE
Radio Communication Systems (RCS) Active Researcher Award, in 2014,
the Asia–Pacific Microwave Conference (APMC) 2014 Prize, and the IEICE
RCS Outstanding Researcher Award, in 2018. He was a co-recipient of the
IEICE Best Paper Award, in 2018, the SoftCOM 2018 Best Paper Award,
and the APCC 2019 Best Paper Award.
VOLUME 8, 2020