Network Aware Spectrum Efficiency Metric For Heterogeneous and Dynamic Radio Environments

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/283125705

Network aware spectrum efficiency metric for heterogeneous and dynamic radio
environments

Article · April 2015


DOI: 10.1109/ACSSC.2014.7094732

CITATIONS READS
2 15

3 authors, including:

Aditya V. Padaki Jeffrey H. Reed


Virginia Polytechnic Institute and State University Virginia Polytechnic Institute and State University
22 PUBLICATIONS   98 CITATIONS    428 PUBLICATIONS   10,870 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Google Study for the Navy View project

LTE in the Unlicensed Spectrum View project

All content following this page was uploaded by Aditya V. Padaki on 18 August 2017.

The user has requested enhancement of the downloaded file.


Network Aware Spectrum Efficiency Metric for
Heterogeneous and Dynamic Radio Environments
Aditya V. Padaki Ravi Tandon Jeffrey H. Reed
Wireless@VT, Dept. of ECE, Discovery Analytics Center, Dept. of CS, Wireless@VT, Dept. of ECE,
Virginia Tech, VA USA Virginia Tech, VA USA Virginia Tech, VA USA
avpadaki@vt.edu tandonr@vt.edu reedjh@vt.edu

Abstract—In this paper, we formalize a new definition for


spectrum efficiency, with the specific goal of addressing the
diverse needs and requirements of various technologies and
users. Existing metrics for spectrum efficiency do not account
for changing radio environments. Hence, they are insufficient
for future systems which employ dynamic allocation schemes.
We introduce a heuristic parameterized definition for spectrum
efficiency dependent on the network dynamics, radio environment
Fig. 1. Disparate Network Dynamics of Shared Spectrum Spaces
and diverse requirements of technologies. This metric accounts
for the frequency use and reuse, interference footprint of a user,
and has a parameter to specify priority/importance for users/bits
(e.g. public safety). We study the behavior and the optimal
transmit power to maximize the efficiency for a point-to-point The traditional measure of ‘bps/Hz’ for spectrum utilization
link. We numerically evaluate the efficiency for a two user case, has several shortcomings. It does not consider the reuse of
and discuss the trade-offs involved. Overall, the proposed metric spectrum, and thus the efficiency monotonically increases with
accounts for the interplay between users in a network to evaluate increasing transmit power. To overcome this, the concept of
efficiency of spectrum utilization.
Area Spectrum Efficiency (ASE) was first introduced in [10]
Index Terms—spectrum efficiency, area spectral efficiency, for cellular systems. Unlike the traditional measure, ASE
dynamic spectrum access, spectrum sharing incorporated the reuse factor to measure efficiency and thus
better represented the spectrum use. This was extended for a
I. I NTRODUCTION hierarchical cell structure in [11]. Hierarchical cell structure is
Spectrum is a scare resource and maximizing its utiliza- similar to the small cell heterogeneous networks, where some
tion is of at most importance to meet the ever increasing or all of a macro cell’s frequency channels are shared by
demands in data rate and bandwidth. Spectrum Sharing and the small cells. These formulations largely render efficiency
Dynamic Spectrum Access (DSA) have been identified as key of spectrum use for specific architectures. This was general-
technologies to make efficient use of the spectrum [1] − [6]. ized for arbitrary wireless systems in [12] and was termed
With new spectrum opening up for sharing [7] − [9], diverse Generalized Area Spectrum Efficiency (GASE). A concept
Radio Access Technologies (RAT) and allied services will of threshold interference tolerance power was introduced and
access the same spectrum band in the future. The spectrum efficiency was formulated to be dependent on this parameter.
will be dynamically managed by a database driven centralized Consequently, GASE marks a threshold on the transmit power,
spectrum management system. [8]. The spectrum management which depends on the interference tolerance power. Increasing
system will aim to maximize the utilization of every available the transmit power beyond this threshold decreases the effi-
Hertz. Thus, metrics which quantify spectrum utilization are ciency. This is because, higher transmit power causes more
vital. degradation through interference to other users relative to the
A clear definition of the spectrum efficiency and its mathe- increase in data rate for the given link.
matical formulation are extremely important to formulate the A principle drawback of GASE however, is that the metric
necessary system models to maximize spectrum utilization. penalizes increasing the transmit power irrespective of whether
The metrics quantifying the individual Quality of Service or not other users are actually located in the vicinity. In other
(QoS) for Multi-RAT systems will be diverse. For example, words, assume that no user is located in the area surrounding
the public safety will measure QoS in terms of reliability, a transmitter, apart from the user it is serving. GASE would
WiFi will measure it based on data rates, and the LTE will still limit the transmit power notwithstanding that no other
measure it based on network coverage, and so on. Measures users are actually getting affected. This unnecessarily curtails
which capture the diversity and indicate the effectiveness of the complete exploitation of the spectrum and thus reduces the
spectrum utilization are needed. In this paper, we propose a overall spectrum utilization. In addition to this, other existing
novel unifying metric for spectrum efficiency, which captures metrics (e.g. [13]) do not differentiate between ‘important’ bits
the overall network dynamics in quantifying the effectiveness and ‘unimportant bits’. The priority of services in futuristic
of spectrum usage. shared spectrum spaces will be policy driven and this will

‹,(((  $VLORPDU


dictate importance of some bits over others. For example,
consider the three scenarios presented in Fig. 1. The existing
metrics are not designed to reflect the network characteristics
and thus render the same efficiency in all the three example
cases. In reality, accounting for network dynamics will aid in
the accurate quantification of spectrum utilization. The lack of
awareness in the formulation does not present a holistic picture
of the network level spectrum efficiency. Since conventional
spectrum spaces were primarily dedicated to fixed services,
network awareness was perhaps not necessary. Consequently
the existing works largely offer network independent formu-
lations for spectrum efficiency which render insufficient for
shared spectrum spaces.
The metric governing the shared spectrum spaces should
represent a holistic view of the network in defining spectrum
efficiency. Consequently, it has to accommodate by providing
more resources for ‘important’ bits irrespective of the inter- Fig. 2. Illustrative Heterogeneous Network
ference caused by such nodes. It also should not penalize any
node’s transmit power in the absence of other users in the
vicinity. In other words, the penalty for causing interference We propose the following definition, derived heuristically to
should be dependent on the actual network characteristics and capture the spectrum efficiency, ηk for user k.
its inherent dynamics. In this paper, we propose a novel,
heuristic, network aware spectrum efficiency metric consid- η k = α k Rk (1)
ering the diverse requirements of various technologies and
services. The proposed metric considers the ‘interference foot- where,

print’ of every user accounting for the actual degradation in the ⎪  1  if dthk ≥ dk
⎨ exp d2th −d2k βk δ
network traffic based on network parameters and dynamics. It k
αk = (2)
further provides for a parameter to specify the priority of a ⎪

particular user as dictated by the policy. 1 otherwise
The rest of the paper is organized as follows: Section II Rk = log2 (1 + γk ) (3)
deals with the development of the novel metric; Section III
discusses the network dependent behavior of the proposed Rk is the data rate per unit bandwidth for user k (depends on
metric for a point-to-point link; Section IV analyses the transmit/received power);
optimal transmit power for a point-to-point link; Section V γk is the Signal-to-Interference-Noise-Ratio (SINR) of user k;
discusses the interplay as seen by this metric between two αk is the weighting factor which is proportional to the number
users using the same channel in a network; and Section VI of nodes in the affected ring (shown in Fig. 2) and priority of
concludes the paper. k;
dthk is the distance at which power of transmitter Txk to
II. D EVELOPMENT OF M ETRIC attenuate to Pth ;
dk is the distance between the transmitter k and receiver k
In this section, we present development of the proposed δ is the node (user) density;
metric. Fig. 2 shows a sample model for heterogeneous node βk is a parameter which specifies the priority or importance
distribution. In this paper, the term heterogeneous is used to of user k, 0 ≤ βk ≤ 1;
describe a network with diverse radio access technologies. Pk is the transmit power by receiver k;
The transmitter (Txk ) is serving a user at a distance dk , with PRk = Pk d−nk is the received power;
a power Pk on frequency fk , on a point-to-point link. Let n is the propagation loss constant;
Pth be a threshold power below which all other users do not Let N0 denote the noise power per unit bandwidth.
face any interference. The received power from Txk would We note the following characteristics about the weighting
attenuate to Pth at a distance dthk from the transmitter. Thus factor αk : (a) 0 ≤ αk ≤ 1; (b) dthk → dk , αk → 1,
all users outside the radius dk , but within the radius dthk face since area of affected users collapses to zero, no users suf-
interference from Txk . This prohibits reuse of frequency fk in fer from interference. Hence the higher weightage; and (c)
the area is termed as ‘area of affected users’ in Fig. 2. dthk → ∞, αk → 0, since all users suffer from interference.
We now introduce two parameters δ and βk . δ denotes the Hence, lower weightage. The exponential function is used in
user density i.e., the number of user devices per unit area. the weighting factor as it brings tractability to the analysis.
βk is a parameter to denote priority of node k. Let K be the Heuristically, it also allows for soft decisions in rewarding
total number of users in the network, indexed by k ∈ (1, K). or penalizing, due to its inherent functionality. Moreover,


the function in the current form yields bounded output for
arbitrarily small inputs.
A good link will consume less spectrum to deliver a given
data rate and allow for a higher reuse factor. This depends
on the channel conditions and receiver capabilities. A bad
link will affect more users in the network than a good link,
and consequently the scaling factor αk will penalize for its
performance. The scaling factor αk penalizes the users based
on the number of nodes affected. The quantity (d2thk − d2k )δ
is proportional to the number of nodes affected by the user
k. The density of nodes δ may be described probabilistically
with a density function depending on time, frequency band and
location. This would provide with an estimate for an expected
spectrum efficiency.
All bits are not created equal. The parameter βk ∈ (0, 1) Fig. 3. Behavior of the proposed metric
is used to specify the priority or importance of one user over
the other. It is inversely proportional to the actual priority,
i.e, βk = 0 indicates highest priority and βk = 1 indicates the same with varying transmit power of Pk . We further vary
least priority. For services which require more priority (e.g. the density of nodes and note the behavior of the proposed
public safety), βk for such user(s) can be set to zero, thus spectrum efficiency metric for user k, with a constant path loss
irrespective of the number of nodes getting affected in the exponent n = 2.8 and distance dk = 1500 m. The resulting
network, reliability and availability of spectrum is not affected curves are shown in Fig. 3. We assume Pth as 10 dB above
for such users. The metric then reduces to the classic definition the nominal noise floor of −165 dBm/Hz, and assume that
of bits per second per unit bandwidth, which is given by interference is absent. The primary interest is to study the
Shannon’s capacity. behavior of this metric from a network management point of
This metric takes into account the network parameters and view.
dynamics, link quality, device capabilities, and QoS require- From Fig. 3, it is evident that for each value set of network
ment types in defining the spectrum efficiency. An interesting parameters (node density, priority, etc.), a unique transmit
parameter in this definition is the threshold power, Pth , whose power maximizes the efficiency. Unlike the traditional metric,
value should be carefully defined. This may be defined either it does not monotonically increase. As can be seen, the optimal
by a policy or a DSA standard (e.g. IEEE 1900.5) as a function transmit power increases with decreasing node density. The
of node density and average signal-to-interference-ratio, or can presence of an optimal transmit power indicates that the metric
be dynamically decided by the spectrum management system accounts for network parameters. So the proposed metric
in real time based on node density. Thus, this metric offers allows the transmitter to choose an optimal power depending
flexibility for spatio-temporal adaptation in accordance with on the network parameters in a dynamic spectrum access
the dynamics of radio environment. environment. This in turn allows for higher frequency reuse,
which increases the spectrum utilization.
A. Network Level Spectrum Efficiency
The spectrum efficiency of each user in the network is given IV. O PTIMAL T RANSMIT P OWER
by (1). However, to ensure the best spectrum utilization, the The optimal transmit power, Pk∗ for a given set of network
spectrum management system has to carefully optimize over parameters for a point-to-point link can be found out by
all the users in the network. It has to account for the interplay maximizing (1) with respect to the transmit power,
and tradeoff between them. It is thus useful to define a network
⎧ ⎫
level spectrum efficiency measure, η given by, ⎪ ⎪

⎨ log2 1 + No P k d −n


K ∗
Pk = max  2   (5)
η= ηk = α k Rk (4) Pk ⎪⎪ Pk n ⎪

⎩ exp Pth
2
− dk βk δ ⎭
k=1 k=1

where αk and Rk are given by (2) and (3) respectively. The closed form solution of this maximization problem can
III. B EHAVIOR OF THE M ETRIC be approximated as,
In this section, we discuss the behavior of the proposed met- ⎡ ⎤ n2
ric for a point-to-point link using equation (1). In particular, we ⎢ 1 ⎥
present the impact of transmit power on spectrum efficiency in Pk∗ = Pth ⎢
⎣ 

n2  ⎦ (6)
1
comparison with conventional bps/Hz for a path loss channel. βk δ W βk δd2k
Pth
No
In Fig. 2, all users within the radius of dthk are affected by
the transmitter Txk . We compute the efficiency metric and plot where W is the Lambert-W function. This is the optimal


Fig. 4. Variation of Optimal Power with Distance Fig. 5. Variation of Maximum Efficiency with Distance

transmit power which maximizes the efficiency for a point-


to-point link.

A. Variation of Optimal Power and Maximum Efficiency


We now discuss the variation of optimal power and maxi-
mum efficiency with distance for a point-to-point link. Fig. 4
depicts the variation of optimal power with distance. For a
given set of network parameters (node distribution, priority,
etc), the optimal transmit power does not increase with dis-
tance at the same rate as the loss of signal power. However,
a sharp increase in the transmit power is seen when the node
density decreases (and similarly for increase in priority). The Fig. 6. Multi-RAT two user network considered
formulation of the metric is such that it facilitates reuse.
Fig. 5 shows the variation of maximum spectrum efficiency
with distance. The efficiency falls sharply with increasing Since interference affects SINR, there is an interplay between
distance of the link. This is because the formulation does not the transmit powers, network efficiency, and individual user
allow for the transmit power to compensate for the signal loss. efficiencies.
This in turn gives an insight for network level management For numerical evaluation, we consider a simple case, where
of spectrum, especially in heterogeneous and shared spectrum d1 = d2 = d = 1500 m and the distances from interfering
spaces. transmitters also is d. The node density chosen is δ = 0.01
per sq. m. In contrast to the single user point-to-point link, the
V. T WO U SER C ASE : F REQUENCY R EUSE AND interference power and nominal noise power are accounted for
C O - CHANNEL I NTERFERENCE SINR calculation. The transmit powers are varied to encom-
In the preceding sections, we addressed only a single point- pass the entire set, and the resulting efficiency for various
to-point link and ignored the frequency reuse factor along values of transmit powers is shown in Fig. 7.
with the resulting co-channel interference. In this section, we The interplay between the two transmitters is clearly ex-
numerically evaluate the efficiency of a two user network, hibited in the figure. Evidently, each user maximizes its own
whose architecture is shown in Fig. 6. We assume a Multi- efficiency when the interfering transmitter is practically turned
RAT heterogeneous network, typical to that of futuristic shared off, as long as the users stay within the realm of interfering
spectrum spaces. We also assume that each of the users are lo- transmitter. This is due to the absence of any interference can-
cated within the interference radius of the other transmitter. We cellation. This gives further insight for network management
further assume that cross technology interference cancellation with Multi-RAT technologies.
is not available and thus, co-channel interference is treated as Fig. 8 presents the observations of a Multi-RAT two user
noise. The distances and transmit powers are denoted by d1 , network graphically. The individual efficiencies are maximized
d2 , P1 , and P2 respectively. The network spectrum efficiency with transmit powers calculated from (5) with no interference
is given by, η = η1 +η2 where η1 and η2 are obtained from (1). from the other transmitter. In other words, maximum efficiency


offers new a perspective on the network level optimization of
wireless systems and aids in the efficient use of spectrum -
especially futuristic networks with Multi-RAT, heterogeneous,
and dynamic spectrum access systems. The trade-offs for a two
user case were also discussed which brought new insights on
achievable regions for spectrum efficiency. Thus, we conclude
by stating that the proposed metric can potentially offer a
different outlook on the quantification of spectrum utilization,
and would serve useful in extracting the maximum of every
available Hertz of spectrum in heterogeneous and dynamic
radio environments.
Future work includes analysis on the selection of threshold
Fig. 7. Network spectrum efficiency for a two user case with interference power for interference tolerance, considering the aggregate in-
terference power from all transmitters in a channel. Evaluation
of the proposed metric for fading channels and hierarchical
network architectures would also provide useful insights for
futuristic wireless systems.
ACKNOWLEDGMENTS
This material is based upon the work supported partly by
National Science Foundation under grant no. EARS-1247928.
Any opinions, findings and conclusions are those of the authors
and do not necessarily reflect the views of NSF.
R EFERENCES
[1] Presidents Council of Advisors on Science and Technology (PCAST)
report, Realizing the Full Potential of Government-Held Spectrum to
Spur Economic Growth, July 2012.
[2] FCC Spectrum Policy Task Force report, Report of the Spectrum
Efficiency Working Group, November 2002.
[3] J. M. Chapin and W. H. Lehr, “The Path to Market Success for Dynamic
Spectrum Access Technology,” IEEE Communications Magazine, May
2007, pp. 96-103.
[4] S. K. Jayaweera, G. Vazquez-Vilar and C. Mosquera, “Dynamic Spec-
trum Leasing: A New Paradigm for Spectrum Sharing in Cognitive
Fig. 8. Graphical representation of observations from a two user case Radio Networks,” IEEE Transactions on Vehicular Technology, vol.59,
no.5, pp.2328-2339, Jun 2010.
[5] I. F. Akyildiz, W. Y. Lee, M. H. Vuran and S. Mohanty, “NeXt
is achieved only with division of radio resources (time or generation/dynamic spectrum access/cognitive radio wireless networks:
A survey,” Computer Networks vol. 50, no. 13, pp. 2127-2159, Sept.
frequency). Moreover, the trade-off for achieving an arbitrary 2006
efficiency within the region depends on the network config- [6] J. H. Reed, J. T. Bernhard, J. M. Park, “Spectrum Access Technologies:
urations. In order to maximize the network efficiency while The Past, the Present, and the Future,” Proceedings of the IEEE, vol.100,
no. Special Centennial Issue, pp.1676-1684, May 2012 (invited paper).
serving all users, the spectrum management system may utilize [7] FCC Notice of Proposed Rule Making, Enabling Innovative Small Cell
the quantification provided by the proposed metric in spectrum Use In 3.5 GHz Band, FCC 12-148, GN Docket No. 12-354, December
allocation algorithms depending on the topology. Thus, the 12, 2012.
[8] FCC Public Notice, 3.5 GHz Licensing Framework, FCC 13-144, GN
proposed metric offers valuable insights for efficient spectrum Docket No. 12-354, November 01, 2013.
use in heterogeneous and dynamic spectrum access networks. [9] FCC Further Notice of Proposed Rule Making, Proposes Creation of
New Citizens Broadband Radio Service in 3.5 GHz, FCC 14-49, GN
VI. C ONCLUSIONS Docket No. 12-354, April 23, 2014.
[10] M. S. Alouini and A. J. Goldsmith, “Area Spectral Efficiency of Cellular
In this paper, we proposed a novel network aware spectrum Mobile Radio Systems,” IEEE Transactions on Vehicular Technology,
efficiency metric which can be used to evaluate the effective- vol. 48, no. 4, pp. 1047-1066, July 1999.
ness of spectrum utilization. The proposed metric accounts for [11] Y. Kim, T. Kwon and D. Hong, “Area Spectral Efficiency of Shared
Spectrum Hierarchical Cell Structure Networks,” IEEE Transactions on
node density, network topology, link quality, device capabili- Vehicular Technology, vol. 59, no. 8, pp. 4145-4151, October 2010.
ties, and user priority of nodes in the network. Area efficiency [12] L. Zhang, H. C. Yang and M. O. Hasna, “Generalized Area Spectral
is implicitly incorporated in the proposed metric. For a point- Efficiency: An Effective Performance Metric for Green Wireless Com-
munications,” IEEE Transactions on Communications, vol. 62, no. 2,
to-point link, the efficiency is maximized for a unique transmit pp. 747-757, 2014.
power, which depends on network parameters. We showed [13] P. Marshall, Quantitative Analysis of Cognitive Radio and Network
through analysis and several numerical results that our metric Performance, Norwood, MA: Artech House, 2010.



View publication stats

You might also like