Content uploaded by B. Walke
Author content
All content in this area was uploaded by B. Walke
Content may be subject to copyright.
488 IEEE COMMUNICATIONS LETTERS, VOL. 15, NO. 5, MAY 2011
User in the Loop: Mobility Aware Users Substantially Boost
Spectral Efficiency of Cellular OFDMA Systems
Rainer Schoenen, Halim Yanikomeroglu, and Bernhard Walke
Abstract—Wireless cellular networks perform with a system
spectral efficiency which depends on the user terminal distribu-
tion over the cell area. Due to the adaptive modulation and coding
schemes used which depend on the signal-to-interference+noise
(SINR) ratio, the achievable data rate is typically an order of
magnitude higher in the cell center compared to the cell edge. The
performance of IMT-Advanced cellular radio systems like IEEE
802.16m and 3GPP LTE-A will strongly depend on algorithms to
cope with the low SINR in the service area. The new paradigm
introduced in this paper motivates users to opportunistically
change location according to operator recommendation displayed
on the user terminal to achieve a much better SINR than
currently available. Benefits are the increase of network capacity
and higher data rates or potentially a financial incentive for
the convinced users. Numeric results based on analysis of IMT
scenarios are provided suggesting large cell spectral efficiency
gains.
Index Terms—IMT-Advanced, spectral efficiency, mobility,
relays, user-in-the-loop (UIL).
I. INTRODUCTION
RECENTLY the IMT-Advanced system performance eval-
uation has been finalized [1]. The requirements by the
ITU [2] are ambitious, therefore various advanced techniques
must be considered, e.g., MIMO, CoMP, and fractional fre-
quency reuse. Multihop techniques (using decode-and-forward
relay nodes, RN) have also been standardized to either increase
cell edge capacity, coverage, or both [3], [4].
The system cell spectral efficiency is the stationary achiev-
able rate averaged in time and over the whole cell area
and normalized by the system bandwidth. This corresponds
to a scheduler in the base station (BS) which distributes
resources fairly among competing user terminals (UTs). When
the goal is to fairly give each UT the same data rate, the
averaging must be calculated differently [5]. In this case a
UT at the cell edge consumes many times more resources
compared to a UT in the cell center. The factor, denoted here
as 𝐹, is given by the ratio of the highest to lowest spectral
efficiency 𝛾in b/s/Hz on a resource block determined by the
adaptive modulation and coding scheme (AMC); for instance,
in Table I, 𝐹=7.5. Especially the AMC is the reason why
the performance depends on the location and distribution of
Manuscript received October 27, 2010. The associate editor coordinating
the review of this letter and approving it for publication was O. Dobre.
R. Schoenen is with the Department of Communication Networks (Com-
Nets), Faculty 6, RWTH Aachen University, Germany, and with the Depart-
ment of Systems and Computer Engineering, Carleton University, Canada
(e-mail: rs@comnets.rwth-aachen.de).
H. Yanikomeroglu is with the Department of Systems and Computer
Engineering, Carleton University, Canada (e-mail: halim@sce.carleton.ca).
B. Walke is with the Department of Communication Networks
(ComNets), Faculty 6, RWTH Aachen University, Germany (e-mail:
walke@comnets.rwth-aachen.de).
Digital Object Identifier 10.1109/LCOMM.2011.042511.102057
TAB L E I
PHY MODES AND SINR INTERVALS
Index 𝑚1 2 3 4 5 6 7 8
SINR 𝜎[dB] 0.9 2.1 3.8 7.7 9.8 12.6 15.0 18.2
Modulation QPSK 16-QAM 64-QAM
Coding rate 1/3 1/2 2/3 1/2 2/3 5/6 2/3 5/6
𝛾[b/s/Hz] 2/3 1 4/3 2 8/310/3 4 5
user terminals in the cell. This factor becomes even worse if
MIMO and advanced receiver algorithms are used.
Operators market their services as if it has, ubiqitiously,
the same QoS everywhere in the service area. While this
would be desirable, in reality large QoS variance is observed
depending on the location. An operator has to provide 𝐹times
the resources to cell edge users compared to cell center users
to provide fair capacity share.
For the voice service 𝐹has to be accounted for when pro-
visioning resources for the busy hour. Further, the maximum
number of simultaneous phone calls depends on 𝐹.
For data services (elastic services, best effort, downloads,
websurfing), the goal of serving each UT with the same
rate (rate-proportional fair service) has the same cost and
dependency on 𝐹as above. At this point it would be more
reasonable to provide each UT a fair amount of resource
blocks (resource-proportional fair service). This is the default
in IEEE 802.11-like systems anyway. The result is indeed a
higher spectral efficiency. The dilemma between throughput
capacity and fairness is known quite well [4].
In this paper we propose a new approach to substantially
increase the spectral efficiency without changing the physical
layer. It begins with the awareness of the user, that the cellular
performance depends on the location. Currently users are
aware of this fact only in IEEE 802.11 hotspot scenarios.
Second, an incentive is needed to improve SINR by moving
the UT to another location. Third, the new location must be
convenient to reach (e.g., on foot) or the incentive must be
enough to make a move. Fourth, there must be an information-
assisted guidance on the UTs, showing directions or even
a map of the area (Fig. 1). This operator database is filled
by all UTs over all times, so it is very substantial. As a
result, some users would be motivated to move to a location
with better SINR and, accordingly, will contribute with a
factor of typically two to four (up to 𝐹) to the increase of
the system’s spectral efficiency. The proposal is opportunistic
(not mandatory), and users do not need to comply with, for
example, when driving in a car.
The method is studied with analytic and numeric tools
through the ITU-R IMT-Advanced standard test scenarios. Our
results show that a substantial gain in cell spectral efficiency
can be achieved with some reasonable effort required by the
users.
1089-7798/11$25.00 c
⃝2011 IEEE
SCHOENEN et al.: USER IN THE LOOP: MOBILITY AWARE USERS SUBSTANTIALLY BOOST SPECTRAL EFFICIENCY OF CELLULAR OFDMA SYSTEMS 489
Fig. 1. Exemplary UT display showing distance, direction and benefitΔ𝑢1,2.
If this is attractive, the user moves to a location with higher spectral efficiency
(with probability 𝑝𝑀).
Fig. 2. Closed loop with the user as the system to control.
II. SYSTEM MODEL
In this paper we assume the IMT-Advanced system
model [2]. Table II provides the data for the scenarios taken
into account. They are representative for the whole range be-
tween densely populated (UMi) to countryside setups (RMa).
Table III specifies parameters related to the LTE-Advanced
radios. The pathloss is calculated using the two parallel models
(with/without line-of-sight) and distance-dependent probabil-
ity 𝑝𝐿𝑂𝑆 (𝑟)to select which one is used [6], [7]. Shadowing
is not modeled here. In the presence of shadowing returns
are expected to be even more favorable. A single antenna
setup is assumed, as the main implications are not affected by
the presence or absence of MIMO. In a multicellular context
with reuse-1 interference is the major limitation. At the cell
borders SINR is close to zero dB with high fluctuations. The
system model includes optional relay nodes (RN) as well
(0 or 3 RN per cell). It is important to select the serving
station (BS or RN) by the least resources decision, i.e., the
decision of taking the single or multihop route is taken by
considering which option uses less resources, not by choosing
𝑚𝑎𝑥(SINR). Over the cell area SINR (𝜎) results are obtained
by numeric analysis and are translated to spectral efficieny 𝛾
in b/s/Hz [8] according to Table I and the modulation and
coding performance results [9].
III. USER IN THE LOOP
The new concept provides suitable information to the user,
e.g., as shown in Fig. 1, and the user should be convinced
TAB L E I I
IMT-ADVANCED SCENARIO SPECIFICATIONS
Scenario Urban Urban Suburban Rural
micro macro macro macro
UMi UMa SMa RMa
Inter-BS distance 200 m 500 m 1299 m 1732 m
BS height 10 m 25 m 35 m 35 m
Antenna tilt −12 ∘−12 ∘−6∘−6∘
𝑓𝐶2.5 GHz 2.0 GHz 2.0 GHz 0.8 GHz
Tx power 44 dBm 49 dBm 49 dBm 49 dBm
TABLE III
TECHNOLOGY PARAMETERS ACCORDING TO LTE-A
Bandwidth (FDD) 20 MHz DL
Trafficfull load; best effort
Antenna gain (boresight) 17 dBi
Sectors/cell 3
Antenna aperture horizontal 𝜃3𝑑𝐵 70 ∘
Antenna aperture vertical 𝜙3𝑑𝐵 15 ∘
Thermal noise -174 dBm/Hz
UT noise figure 5dB
to change his location voluntarily from his current location
⃗𝑝1=(𝑥1,𝑦
1)to ⃗𝑝2. Thus the user becomes part of a control
loop (Fig. 2). System theory including human elements is
inspired by [10]; power supply companies have been trying out
such approaches [11] in recent years. The network controller
knows the current signal quality 𝜎(⃗𝑝1)(SINR-based) or 𝛾(⃗𝑝1)
from UT measurements, and the expected level 𝛾(⃗𝑝2)from
a database of measurements of all UTs at all locations in
the past. The user knows his utility advantage of Δ𝑢1,2=
𝑢(⃗𝑝2)−𝑢(⃗𝑝1)when doing the move. This utility 𝑢can be either
financial (savings for voice during busy hours) or an increased
data rate (for best effort data traffic). The network provides the
information in which direction or to which location to move
by the gradient −▽𝜎(⃗𝑝)of the potential field at position ⃗𝑝1.
The user should have all information to make his decision. UT
devices would ideally have GPS onboard, but in the absense
of GPS the network can still support ranging by BS-based
triangulation and give hints for movement. The user can see
in which direction to move best and how far 𝑑=∣⃗𝑝2−⃗𝑝1∣
the next improvement step is. It is assumed that a fraction
𝑝𝑀of users actually participates in moving, the rest stay at
place. 𝑝𝑀accounts for users that cannot move, do not want to
move, or have no sufficient incentive to move. The output of
the user block (Fig. 2) is the new location ⃗𝑝2. It is described
by a Bernoulli random process where 𝑝𝑀is the probability
of a move from ⃗𝑝1to ⃗𝑝2for 𝑑meters and (1 −𝑝𝑀)of no
movement at all. The target value 𝛾Θis the least 𝛾this UT
should achieve after the movement.
IV. PERFORMANCE RESULTS
In this paper the statistics of the movement distance and
the resulting spectral efficiency are determined by numeric
analysis of the IMT-Advanced scenarios. Both increases of Δ𝛾
and 𝑑are weighted by 𝑝𝑀, because (1 −𝑝𝑀)of the users are
assumed to be not willing to move. Users who do not need to
move as they are already at good positions are accounted with
𝑑=0. Rather moderate values for the parameters have been
chosen in order to be realistic: 𝑝𝑀=1
2and 𝛾Θ=2.5b/s/Hz.
The scenarios were investigated with and without relays.
490 IEEE COMMUNICATIONS LETTERS, VOL. 15, NO. 5, MAY 2011
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
1
1.5
2
2.5
3
3.5
4
4.5
5
target (threshold) value [bits/s/Hz]
Cell spectral efficiency [bits/s/Hz]
pM=0.0
pM=0.1
pM=0.2
pM=0.3
pM=0.4
pM=0.5
pM=0.6
pM=0.7
pM=0.8
pM=1.0
pM=0.9
Fig. 3. Observed spectral efficiency 𝛾in b/s/Hz depending on the reference
𝛾Θin the SMa scenario.
TAB L E I V
AVERAGE SPECTRAL EFFICIENCY AND AVERAGE DISTANCE RESULTS FOR
THE IMT SCENARIO EVALUATION [B/S/HZ/SECTOR]WITH 𝛾Θ=2.5
B/S/HZ
Scenario UMi UMa SMa RMa
0RN,HUD ¯𝛾1.567 1.254 1.234 1.974
0RN,UIL ¯𝛾2.170 1.995 2.836 2.509
𝑝𝑀=0.5¯
𝑑4.4 m 4.7 m 7.8 m 30.7 m
0RN,UIL ¯𝛾2.772 2.735 4.437 3.045
𝑝𝑀=1 ¯
𝑑8.8 m 9.4 m 15.6 m 61.4 m
3RN,HUD ¯𝛾1.945 1.804 1.825 2.310
3RN,UIL ¯𝛾2.333 2.239 2.858 2.654
𝑝𝑀=0.5¯
𝑑1.9 m 1.7 m 5.0 m 12.0 m
3RN,UIL ¯𝛾2.721 2.674 3.892 2.998
𝑝𝑀=1 ¯
𝑑3.7 m 3.4 m 10.0 m 23.9 m
The spectral efficiency and average distance results for the
IMT-Advanced scenarios are provided in Table IV, where
HUD means homogeneous user distribution (classic conserva-
tive model without movement [7]) and UIL means user-in-the-
loop (progressive model) with anisotropic, nonhomogeneous
user density. The average cell spectral efficiency is increased
by 25% to more than 100%, depending on the IMT scenario,
even with moderate parameter values of participation 𝑝𝑀and
threshold 𝛾Θ. This increase requires an effective movement of
just a few meters on average. The parameter 𝑝𝑀influences
the result linearly, whereas 𝛾Θdoes not (Fig. 3).
Figure 4 shows the cumulative density function (CDF) of
the required movement. It explains how many users have to
move by how far. In all scenarios (except RMa) 80% to 90% of
the users need to move less than 10 m. Relays reduce the effort
to move for the user due to the shorter distance to the closest
RN. If shadowing is present even higher returns are expected.
The analysis results suggest that the proposed approach works
well, because the required movement is small, while the gains
are high. Results for other scenarios can be found in [12].
If implemented, user acceptance is expected to increase with
time and training.
V. C ONCLUSION
This paper introduced a novel approach to increase the
system spectral efficiency by the help of participating users. It
is not that these users are interested in this goal itself [13], but
they have an incentive to voluntarily improve their application
data rate or reduce the cost of connections during busy hours
by performing a voluntary movement to a location of higher
0 102030405060708090100
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Distance d to move in meters
Cumulative Distribution Function CDF(d)
RMa,0RN
RMa,3RN
SMa,0RN
SMa,3RN
UMa,0RN
UMa,3RN
UMi,0RN
UMi,3RN
Fig. 4. CDF of the movement distance 𝑑for all IMT scenarios. 0and 3in
the scenario name parameter mean the number of relay nodes in the cell.
performance. Studies using the IMT-Advanced evaluation sce-
narios showed substantial gains up to 200%, depending on the
percentage of users involved. The distances to move are in the
order of a few meters in most cases. It is recommended that
the user-in-the-control-loop techniques should be investigated
further, as this approach seems to be promising to tackle many
economic and ecological problems.
Future work will contain results of a recent user survey,
green index motivation and financial incentives or dynamic
tariffs to avoid mobile data traffic congestion in busy hours.
REFERENCES
[1] J. Monserrat et al., “Advanced radio resource management for IMT-
Advanced in WINNER+ (II),” in Proc. ICT-MobileSummit, June 2010.
[2] ITU, “Report ITU-R M2135; guidelines for evaluation of radio interfer-
ence technologies for IMT-Adcanced,” 2008.
[3] R.Pabst,B.Walke,D.C.Schultz,H.Yanikomerogluet al.,“Relay-
based deployment concepts for wireless and mobile broadband radio,”
IEEE Commun. Mag., pp. 80–89, Sep. 2004.
[4] M. Salem, H. Yanikomeroglu, D. Falconer et al., “An overview of ra-
dio resource management in relay-enhanced OFDMA-based networks,”
IEEE Commun. Surveys and Tutorials, vol. 12, no. 3, 2010.
[5] C. Hoymann and S. Goebbels, “Dimensioning cellular WiMAX part I:
singlehop networks,” in Proc. European Wireless, Apr. 2007, p. 7.
[6] D. B¨ultmann, T. Andre, and R. Schoenen, “Analysis of 3GPP LTE-
Advanced cell spectral efficiency,” in Proc. PIMRC, Sep. 2010.
[7] R. Schoenen and C. Teijeiro, “System level performance evaluation of
LTE with MIMO and relays in reuse-1 IMT-Advanced scenarios,” in
Proc. IEEE WiCom, Sep. 2010.
[8] K. Brueninghaus and D. e. a. Astely, “Link performance models for
system level simulations of broadband radio access systems,” in Proc.
PIMRC, Sep. 2005, pp. 2306–2311.
[9] R. Schoenen and B. Walke, “On PHY and MAC performance of 3G-
LTE in a multi-hop cellular environment,” in Proc. IEEE WiCom,Sep.
2007.
[10] R. Zimmerman, Das System Mensch. Artech House, 2003.
[11] A. Faruqui and R. Earle, “Demand response and advanced metering,”
Regulation, vol. 29, no. 1, pp. 24–27, 2006.
[12] R. Schoenen, “On increasing the spectral efficiency more than 100%
by user-in-the-control-loop,” in Proc. 16th Asia-Pacific Conference on
Communications, Oct. 2010.
[13] G. Hardin, “The tragedy of the commons,” Science, vol. 20, pp. 1243–
47, 1968.
