The emerging smart utility networks (SUN) provide twoway communications between smart meters and smart appliances for purpose of low power usage, low cost, and high reliability. This paper deals with a bandwidthefficient communication method based on the hidden pilotaided scheme using a precoder in downlink SUN suitable for highrate multimedia applications. With the aid of the design of a precoder and a superimposed hidden pilot, it is possible to estimate the channel without loss of bandwidth. In the channel estimation procedure, the inevitable data interference, which degrades the performance of channel estimation, can be reduced by the precoder design with an iterative scheme. Computer simulations show that the proposed scheme outperforms the conventional method in terms of achievable data rate, especially when a large number of subcarriers are employed.
1. INTRODUCTION
The increasing interest in modern smart grid requires a new technology for providing the twoway communication between the consumer and the utility at the appliance/meter level
[1

6]
. This will enable the consumer to continuously monitor power usage and make instantaneous decisions to control electric devices at consumers’ homes or offices to reduce energy usage, cost and increase service reliability. Smart utility networks (SUN), a subordinated network in the smart grid, consist of smart home appliances and meters equipped with radio terminals. Smart devices in SUN can automatically control power consumption during peak and offpeak load times according to the demand requirement. This will ensure that power is not utilized by the appliance when the cost of power consumption is expensive. In addition, SUN may support other functionality needs, such as multimedia data transmission from substations to smart devices in addition to normal control data.
IEEE 802.15.4g specification for Physical Layer (PHY) is built for smart devices of SUN
[7]
. Among the various alternative PHY designs, multirate and multiregional orthogonal frequency division multiplexing (MROFDM) was designed to provide higher data rates in channels that have significant delay spread. Due to the fast growth of demands on multimedia services of smart devices, SUN are expected to support high data rate applications. In addition, in order to reduce the power consumption and interference during communications between smart devices, a transmission method to shorten transmission time is required. To solve these problems, multipleinput multipleoutput (MIMO) systems can be a viable solution for high data rate communications over SUN. As the downlink of the SUN with dense smart meters will act as the bottleneck link, an efficient usage of downlink wireless resources is important to provide high quality of communication services.
In the wireless SUN, a pilotsymbolaided modulation (PSAM) scheme, one of the most widely used schemes in cellular systems, can be used to estimate the channel state information (CSI) exploiting pilot symbols that are previously known at the user terminal
[8]
. Accurate channel estimates can be obtained by using the judiciously placed pilots
[9]
, whereas it consumes bandwidth which is valuable in wireless communications. Although blind channel estimation
[10]
offers a bandwidthefficient method exploiting statistical and other properties of the data, it has its inherent shortcomings, such as slow convergence and phase ambiguity. An alternative method is the superimposed training (ST) scheme
[11

12]
has a virtue of no loss in information rate at the expense of data power. However, this method may degrade the performance of channel estimation due to the effect of inherent interference of data. Use of a hidden pilot based precoding scheme can be a feasible solution to this problem, as it removes inevitable unknown data interference with the aid of the design of a precoding and a ST sequence
[13]
.
In this paper, we propose a bandwidth efficient precoding scheme for downlink SUN for achieving high data rates. Considering different designs of a precoder and a superimposed hidden pilot per transmit antenna in smart devices, spatially multiplexed data can be successfully recovered. In addition, due to the spreading effect of the information data over the total available bandwidth using a precoder, it is possible to provide high frequency diversity. The performance gain of achievable data rate becomes significant when a large number of subcarriers is employed.
This paper is organized as follows. In Section 2, a system model of the proposed scheme is introduced. Section 3 presents the channel estimation and symbol detection. The results of computer simulations are shown in Section 4. Finally, conclusions are drawn in Section 5.
2. SYSTEM MODEL OF A HIDDEN PILOTAIDED PRECODING SCHEME
 2.1 MROFDM system model
We consider a downlink transmission from an access point, such as a coordinator or a router, equipped with two transmit antennas, to
K
smart devices, such as inhome and metering devices, also having two receive antennas, depicted in
Fig. 1
. At each transmit antenna, a MROFDM modulator followed by the proposed hidden pilot aided precoder is used. Let
N
be the total number of subcarriers in one OFDM symbol. Remind that application of block transmission with a cyclic prefix (CP) over a timedispersive channel can be modeled as transmission over a channel with a circulant channel matrix. And the length of the CP
L
is set to be equal to the number of nonzero taps of the impulse response for each channel.
The downlink smart utility network.
After removing the CP at the
k
’th smart device’s
i
th receive antenna, we obtain an
N
×1 vector
which can be written as
where
is an
N
×1 precoded transmit symbol block for the
k
th assigned device, from the
j
th transmit antenna.
is the
N
×
N
downlink channel matrix, which is a circulant matrix with the first column given by
is the
L
×1 vector representing the length
L
channel impulse response from the coordinator’s
j
th transmit antenna to the
i
th receive antenna of the
k
’th smart device. The
N
×
N
discrete Fourier transform (DFT) matrix
F
_{1}
has an (
m
,
n
) entry
m,n
=0,1,...,
N
–1. We assume that
is additive white Gaussian noise (AWGN), whose elements are independent and identically distributed (i.i.d.) zero mean complex Gaussian random variables with variance
 2.2 Design of a precoder and a superimposed hidden pilot
In this subsection, we derive the design of a proposed scheme based on a precoder and a ST symbol. Since a training sequence for channel estimation is simply added to the precoded data, there exists mutual interference between them. To mitigate inherent mutual interference, the precoder based on orthogonal polyphase sequences
[13]
, which have good periodic autocorrelation and crosscorrelation properties, is exploited. These properties are utilized to eliminate the interference between the data and the hidden pilot guaranteeing almost no loss of transmit data rate. To generate these sequences, the length of a sequence is set to
p
^{r–1}
, where
p
is prime and
r
(>1) is an integer. Then, the
N
/2
K
–1 (=
p
^{r–1}
) numbers of polyphase sequence sets such that
C
=[
c
_{0}
,
c
_{1}
,...,
c
_{N}
_{/2}
_{K}
_{–2}
], where
c_{i}
= [
c_{i}
(0),
c_{i}
(1),...,
c_{i}
(
N
/2
K
–2)]
^{T}
, can be generated by
Then, we can split the polyphase sequence sets into two parts, the sequence for a precoder and for a hidden pilot. Design of the precoder matrix
P
and the pilot vector
t
can be expressed as follows:
where
c
’=[
c^{T,}
0]
^{T}
is an
N
/2
K
×1 vector and the
N
/2
K
×
N
/2
K
discrete Fourier transform (DFT) matrix
F
_{2}
has an (
m,n
) entry
m,n
=0,1,...,
N
/2
K
–1.
We design the transmit symbol block using the above precoder and hidden pilot with regard to the subcarrier and the antenna index. To support the orthogonality of different antennas’ pilot signals for channel prediction, we consider disjoint sets of tones where hidden pilot symbols are included or not in transmit symbol at each antenna. First, we set the
N
/2
K
×
N
/2
K
subcarrier index matrix
i_{q}
selecting odd and even index vectors, respectively. Here, we assume that an (
N
/
K
–4)×1 information data block for the
k
th smart device from the
j
th transmit antenna
is the i.i.d. zero mean complex Gaussian random vector with covariance
P_{s}
/(
N
/
K
–4)
I
_{N}
_{/K}
_{–4}
. We also assume that
P_{s}
+
P_{t}
=1 where
P_{t}
denotes the transmit pilot power in one symbol block.
We design the transmit symbol block using the above precoder and hidden pilot with regard to the subcarrier and the antenna index. To support the orthogonality of different antennas’ pilot signals for channel prediction, we consider disjoint sets of tones where hidden pilot symbols are included or not in transmit symbol at each antenna. First, we set the
N
/2
K
×
N
/2
K
subcarrier index matrix
i_{q}
selecting odd and even index vectors, respectively. Here, we assume that an (
N
/
K
–4)×1 information data block for the
k
th smart device from the
j
th transmit antenna
is the i.i.d. zero mean complex Gaussian random vector with covariance
P_{s}
/(
N
/
K
–4)
I
_{N}
_{/K}
_{–4}
. We also assume that
P_{s}
+
P_{t}
=1 where
P_{t}
denotes the transmit pilot power in one symbol block.
As previously mentioned, to locate disjoint sets of tones, which include hidden pilots or not, in the transmit symbol at each antenna, we design the precoded vector
as
where
are the odd and even index information data blocks from
respectively. We then obtain the transmit symbol block
which is shown as
where
Φ_{k}
is a subcarrier mapping matrix which has the following conditions:
Note that we apply equidistantlylocated subcarrier mapping which spreads the signal over the total available subcarriers to obtain frequency diversity as much as possible. It is clear from Eq. (5) that the proposed design of hidden pilots imposes good bandwidth efficiency, because we can locate information data over the available MROFDM subcarriers and pilots are simply added to certain subcarriers determined by
At the
k
’th smart device’s
i
th receive antenna, an
N
/2
K
×1 vector
is selected by the subcarrier demapping matrix
for channel estimation and symbol detection, which is given by
where
is an
N
/2
K
×
N
/2
K
diagonal channel matrix whose principal diagonal components are the elements of
which is the frequency response of a time domain channel
3. CHANNEL ESTIMATION AND SYMBOL DETECTION
 3.1 Channel Estimation
At smart devices using the hidden pilotaided precoding scheme, channel estimation can be carried out by preprocessing received signals in order to reduce the interference between the data and the hidden pilot. The preprocessing is based on the correlation properties of the designed precoder and the hidden pilot. Using the hidden pilot and the inverse DFT matrix for preprocessing, the interference
consisting of data symbols is suppressed as
where
B
is an
N
/2
K
×
L
circulant matrix with the first column
b
,
is an
N
/2
K
×
N
/2
K
circulant matrix with the first column
is a modified noise vector, i.e.,
We have used the fact that
due to the commutativity of circular convolution. If we let
A_{l}
be a columnwise circulant matrix with the
l
th column of
A
as the first column, which we denote as
is equivalent to
As depicted in
Fig. 2
, the polyphase sequence has a good periodic crosscorrelation property. From the figure, we can notice that
B^{H}A_{l}
approaches 0 for all
l
, and thus, we can effectively remove the data interference
Then, we can estimate the time domain channel by the linear MMSE channel estimator
[14]
given by
Periodic crosscorrelation property of the polyphase sequences with kshifted and nshifted vectors.
Now, let us define the channel estimation error as
and the corresponding mean square error (MSE) can be expressed as
where
is the variance of noise
In fact,
B
shows a good periodic autocorrelation feature, i.e.,
B^{H}B
→
_{c}I
where
c
is a constant and
I
is the identity matrix.
 3.2 Symbol Detection
Using the estimated CSI, the received signal vector
is equivalently expressed as
where
denote diagonal matrices whose main diagonals are
respectively. Since we consider smart devices equipped with two antennas, we can combine two received signal vectors into one signal vector, that is,
where
We consider a linear MMSE receiver
[12]
to detect information data from the received symbols. Then, symbol detection is presented as
where
P_{s}
is the total data symbol power in one block. After finishing linear MMSE symbol detection, we can additionally suppress the residual of data interference components by using an iterative algorithm, which further decreases
with the help of the detected symbol and the estimated channel. Carrying out the iterative algorithm, a new preprocessed data block
can be written as
where
is the
N
/2
K
×
N
/2
K
circulant matrix with the first column
It was observed by numerical simulations that the iterative algorithm for accurate channel estimation converges rapidly within a couple of iterations.
4. SIMULATION RESULTS
In this section, the performance of the proposed scheme is evaluated under SUN. For the purpose of comparison, a PSAM scheme is presented as the conventional scheme. As shown in the IEEE 802.15.4g PHY specification
[5]
for SUN, the total subcarrier number
N
in MROFDM is set to 128 for the SUNOFDM option 1 and 64 for the option 2. It is assumed that a coordinator transmits data to two smart devices, where each device is equipped with two antennas. The channel is randomly generated at each MonteCarlo run and is assumed to be Rayleigh with length
L
=8 for OFDM option 1 and
L
=4 for OFDM option 2. Without loss of generality, the number of the disjoint pilot symbols in conventional PSAM is set to the number of channel length.
The channel estimation performance of the proposed scheme based on a hidden pilot is compared to the performance of conventional PSAM in
Fig. 3
. Here, the pilot power ratio loaded to the hidden pilot is set to 0.7 and the SUNOFDM option 1 is considered. Note that the analysis of channel estimation error derived in Eq. (11) is similar to the curve obtained with Monte Carlo simulations. In addition, it is shown that the channel estimation performance of the proposed scheme is better than the PSAM scheme in the low SNR region because the proposed one uses a more sufficient number of pilots compared to PSAM. However, in the high SNR, the channel MSE of the proposed scheme is worse since the fact that
B^{H}A_{q}
→0 is not perfectly satisfied, and thus, it degrades the channel estimation performance. Such a residual interference can be reduced by iteratively updating the detected symbols and the estimated channel, as shown in
Fig. 3
.
Performance comparison of channel estimation of the proposed hidden pilotaided precoding scheme and the conventional PSAM.
Fig. 4
displays the achievable data rate performance of the proposed scheme with regard to the power allocated to the hidden pilot. The power ratio loaded to the hidden pilot varies from 0.3 to 0.7 in the SUNOFDM option 1 scenario. When SNR is low, the performance of channel estimation has a minor effect on the achievable data rate due to the severe communications condition. However, in the high SNR region, the imperfect channel prediction leads to the notable performance degradation. To obtain high achievable data rate for the high SNR case, the use of pilot power
P_{t}
=0.7 is desirable.
Achievable data rate comparison for the SUNOFDM option 1 and option 2.
The comparison of the achievable data rate between the proposed scheme and the conventional PSAM for the SUNOFDM option 1 and option 2 are shown in
Fig. 5
. For both cases, the proposed scheme outperforms the conventional PSAM under overall SNR region. For channel prediction, the proposed scheme does not waste valuable bandwidth, while the PSAM scheme consumes pilot tones which degrades bandwidth efficiency. It can be observed that the proposed scheme provides better achievable data rate in option 1, where a large number of subcarriers are considered, than the one in option 2. Since the precoder in the proposed scheme spreads each symbol over the available frequency domain and achieve a high degree of frequency diversity, the use of a large number of subcarriers, based on the SUNOFDM option 1, results in enhanced performance of achievable data rate.
Achievable data rate for different values of power loaded to the hidden pilot.
5. CONCLUSION
For the bandwidth efficient communication method in smart utility networks (SUN), a hidden pilot based precoding scheme, which mitigates the mutual interference between the data and pilot symbols, is proposed. The downlink configuration consisting of a coordinator and smart devices, equipped with multiple antennas to provide multimedia data transmissions, is considered. To obtain frequency orthogonality for estimating multipleantenna channels, we have designed disjoint sets of tones, which include the hidden pilot or not. Since the proposed scheme does not consume bandwidth for channel estimation and provides high frequency diversity, it yields better achievable data rate performance than the conventional PSAM scheme in SUN. The performance gain becomes significant when the use of a large number of subcarriers, presented as the MROFDM option 1 in PHY SUN standard, is considered.
BIO
Byung Wook Kim
He received the B.E. degree from Pusan National University, Pusan, Korea, in 2005 and the M.S. and Ph.D. degrees in Electrical Engineering from KAIST, Daejeon, Korea, in 2007 and 2012, respectively. From 2012 to 2013, he was a senior engineer of Korea Electrotechnology Research Institute (KERI), Changwon, Korea. Currently, he is working as an assistant professor in the School of Electrical and Railway Engineering at Kyungil University. His research interests are in wireless communications, power line communications, visible light communications.
NIST
2009
The Smart Grid Interoperability Standards Roadmap
2011
IEEE 802. 15. 42011, Part 15. 4: LowRate Wireless Personal Area Networks
Chang K.H.
,
Mason B.
2012
“The IEEE 802.15.4g Standard for Smart Utility Networks”
Proceedings of IEEE 3rd International Conference on Smart Grid Communications (SmartGridComm)
476 
480
Choi D.
,
Chung I.
2013
“A Study of an Efficient ZigBee Address Assignment Scheme for Home Area Networks of Smart Grid”
Journal of Korea Multimedia Society
16
(5)
566 
576
DOI : 10.9717/kmms.2013.16.5.566
Kim J.S.
,
Lee H.Y.
,
Lee Y.R.
,
Kim B.W.
2012
“Smart Social Grid System using Interactive Sketch Map”
Journal of Korea Multimedia Society
15
(3)
388 
397
DOI : 10.9717/kmms.2012.15.3.388
Hart Dr David
2009
Smart Grid Overview, 1509041400wng0
2012
IEEE Standard 802. 15. 4g, Part 15. 4: LowRate Wireless Personal Arae Networks (LRWPANs) Amendment 3: Physical Layer (PHY) Specifications for LowDataRate, Wireless, Smart Metering Utility Networks
Jeon W.G.
,
Paik K.H.
,
Cho Y.S.
2000
“An Efficient Channel Estimation Technique for OFDM Systems with Transmitter Diversity”
Proceedings of IEEE International Symposium on Personal , Indoor and Mobile Radio Communications
Vol. 2
1246 
1250
Adireddy S.
,
Tong L.
,
Viswanathan H.
2002
“Optimal Placement of Training for Frequencyselective Blockfading Channels”
IEEE Transactions on Information Theory
48
(8)
2338 
2352
DOI : 10.1109/TIT.2002.800466
Bolcskei H.
,
Heath R.W.
,
Paulraj A.J.
2002
“Blind Channel Identification and Equalization in OFDM based Multiantenna Systems”
IEEE Transactions on Signal Processing
50
(1)
96 
109
DOI : 10.1109/78.972486
Zhou G.T.
,
Viberg M.
,
McKelvey T.
2003
“A Firstorder Statistical Method for Channel Estimation”
IEEE Signal Processing Letters
10
(3)
57 
60
DOI : 10.1109/LSP.2002.807864
Ghogho M.
,
Swami A.
2004
“Improved Channel Estimation using Superimposed Training”
Proceeding of International Conference on Acoustics, Speech and Signal Processing
110 
114
Kim B.W.
,
Jung S.Y.
,
Kim J.
,
Park D.J.
2008
“Hidden Pilot based Precoder Design for MIMOOFDM Systems”
IEEE Communications Letters
12
(9)
657 
659
DOI : 10.1109/LCOMM.2008.080681
Proakis J.G.
,
Salehi M.
2008
Digital Communications
McGrawHill Higher Education
New York