Sie sind auf Seite 1von 5

1556 IEEE SIGNAL PROCESSING LETTERS, VOL. 22, NO.

10, OCTOBER 2015

Transceiver Optimization for Unicast/Multicast


MIMO Cognitive Overlay/Underlay Networks
Nikhil Gupta and Aditya K. Jagannatham

AbstractIn this work, we develop a majorization theory based Thus, mean squared error (MSE) minimization, which has been
linear precoding framework for optimal transceiver design in proposed in works such as [10], can be employed as a reliable
MIMO cognitive radio networks. Closed form expressions are alternative approach for performance optimization in practical
derived for the optimal MIMO precoders using two new trans-
cognitive radio scenarios. The work in [11] presents a frame-
ceiver design paradigms, the zero-forcing transceiver (ZFT) and
the interference optimized transceiver (IOT), for overlay and work for MSE minimization in cognitive radio scenarios. How-
underlay MIMO cognitive radio networks respectively. Further, ever, the scheme presented therein is based on a iterative pro-
another novel contribution of this work is to derive the precoders cedure which involves repeated matrix inversion and therefore
for multicast MIMO cognitive radio scenarios based on novel has a high computational complexity. In this context, a frame-
multi-user mean-squared error (MSE) bounds. Simulation results work for optimal MIMO transceiver design towards MSE min-
demonstrate the performance of the proposed optimal MIMO imization, based on majorization theory, has been presented in
transceivers.
[10] and offers an attractive solution for optimal precoder de-
Index TermsCognitive radio, MIMO, transceiver optimiza- sign. However, the work therein cannot be directly applied in the
tion. context of interference constrained cognitive radio scenarios.
More importantly, the work therein is restricted to single user
unicast scenarios. Consequently, in this letter, we propose two
I. INTRODUCTION
new MIMO transceiver design paradigms for cognitive radio

C OGNITIVE radio has recently emerged as a popular new


paradigm towards relieving spectral congestion [1], [2]
by allowing secondary users to conditionally access spectral
networks, i.e. the zero-forcing transceiver (ZFT) and the inter-
ference optimized transceiver (IOT), employing majorization
theory. Depending on the cognitive radio policy, one can either
bands licensed to primary users. Naturally, it is essential for choose ZFT when absolutely no secondary user interference is
the secondary users to limit the interference caused to the pri- allowed or IOT when a marginal level of interference can be
mary wireless users. Several MIMO precoding techniques have tolerated at the primary user. The proposed framework can be
been developed to optimize the secondary transmission. Works employed for transceiver design in diverse cognitive radio sce-
[3], [4], [5] present beamforming algorithms for cognitive radio narios such as overlay, where the secondary transmission causes
scenarios with single antenna users and do not focus on op- no interference to the primary user due to intelligent signal pro-
timal precoders for general multi-antenna users. Further, the cessing and underlay, where a nominal interference level can
proposed algorithms in [3], [4] are iterative in nature, while the be tolerated at the primary user. A novel contribution of this
work in [5] does not present any closed form solution. Optimal work is to present a comprehensive framework for zero-forcing
MIMO precoding schemes such as block diagonalization and and interference threshold based transceiver design for MSE
successive optimization [6], [7] have become signicantly pop- minimization, whereas works such as [8], [12] consider only
ular. The authors in [8] have developed two schemes namely sum-rate optimization. We consider a variety of design objec-
the P-SVD and D-SVD for zero-forcing and interference con- tives such as sum/product MSE minimization and the min-max
strained MIMO cognitive radio transmission respectively, while MSE criteria to derive closed form solutions for the optimal pre-
the authors of [9] present a framework for secondary user rate coders, which have a signicantly lower computational com-
maximization subject to a primary user rate constraint. How- plexity compared to the iterative schemes in works such as [11].
ever, the above works are based on power allocation towards Further, for multicast scenarios arising in 3G/4G networks, pre-
sum rate maximization. Realization of the sum rate performance coder design is challenging due to the simultaneous transmis-
requires a signicant additional complexity in terms of forward sion to several users. In this context, we present novel precoder
error correction (FEC) at the transmitter and receiver, and there- design expressions based on various MSE bounds derived for
fore does not correlate well with the actual BER performance. the sum, product and min-max MSE criteria.

II. OPTIMAL MIMO TRANSCEIVERS FOR UNICAST


Manuscript received December 13, 2014; revised February 12, 2015;
accepted March 10, 2015. Date of publication March 18, 2015; date of current
COGNITIVE RADIO SCENARIOS
version March 24, 2015. The associate editor coordinating the review of this Consider a cognitive radio scenario with primary users, a
manuscript and approving it for publication was Prof. Francesco Verde.
The authors are with the Department of Electrical Engineering, Indian In- secondary transmitter and receiver. Let denote the number
stitute of Technology, Kanpur, UP 208016, India (e-mail: nikgupta@iitk.ac.in; of antennas at the th primary user, while , denote
adityaj@iitk.ac.in). the number of antennas at the secondary transmitter and re-
Color versions of one or more of the gures in this paper are available online
at http://ieeexplore.ieee.org. ceiver respectively. The secondary user channel matrix is de-
Digital Object Identier 10.1109/LSP.2015.2413940 noted by , while the MIMO channel between

1070-9908 2015 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
GUPTA AND JAGANNATHAM: UNICAST/MULTICAST MIMO COGNITIVE OVERLAY/UNDERLAY NETWORKS 1557

the secondary transmitter and the th primary user is denoted is given as , where denotes the diagonal
by . Thus, the received signal matrix with elements of vector along the
at the secondary receiver is given as, principal diagonal, and are given as,

(1)
and the interference at primary user from the secondary user is
given as where is the symbol vector trans-
mitted for the secondary user by the secondary transmitter and where is the Lagrange multiplier such that and
denotes the precoding matrix for the secondary if and 0 otherwise.
Proof: Similar to result in [10]. Follows from the fact that
user. The quantity represents the noise plus pri-
the optimization objective is a Schur concave func-
mary user interference at the secondary user, with covariance
tion. Therefore, the optimal zero forcing precoder which min-
.
imizes the sum MSE is given as where are
A. Zero-Forcing Transceiver Design (ZFT) as given in (1) above.
Consider now the optimal ZFT which minimizes the product
We now begin with a description of the zero-forcing based op-
MSE . This is a Schur-concave objective function and
timal transceiver design scheme ZFT for MIMO cognitive radio
the optimal ZFT precoder which minimizes the product MSE
scenarios. The ZFT nulls the interference at the primary users,
above is given as , where . Fur-
while simultaneously minimizing the MSE of transmission of
the secondary user, thereby optimizing the secondary user per- ther, the optimal ZFT precoder which minimizes the maximum
formance. Consider the concatenated primary user channel ma- MSE i.e. the optimization objective is given as
, where , and the matrix
trix given as,
is any unitary matrix which satises such as the
Hadamard matrix. It can be noted that the ZFT design is pos-
sible only when , which is a stringent con-
dition. The IOT design proposed next however does not require
the above condition to be met and is therefore more exible in
where above denotes the SVD of the matrix . There- terms of practical implementation.
fore, denotes the null space of the concatenated primary user
channel matrix. Similar to works such as [8], [11], let B. Interference Optimized Transceiver Design (IOT)
denote the noise-whitened effective channel ma-
trix of the secondary user and denote the precoding matrix. While the optimal ZFT described above completely cancels
The secondary precoder can be readily seen to null the interference to the primary users, this can lead to a degra-
the interference at each of the primary users. Consider now dation in the performance of the secondary user. However, in
the standard MIMO linear receiver given as, scenarios where the primary users are capable of tolerating in-
terference below a pre-dened threshold, one can use the IOT
based optimal precoder design procedure described below, and
is signicantly different in nature from the work in papers such
as [12], [8] which consider only multiuser performance opti-
where for the zero-forcing and MMSE receivers re- mization. Let the acceptable interference level be denoted by
spectively. Therefore, corresponding to the precoding matrix . Let the SVD of , with denoting the
and the MIMO receiver , the covariance matrix for the error singular values. The net interference to the primary users corre-
at the receiver can be seen to be given as, sponding to the precoder is given as,

Let denote the th diagonal element of the covariance matrix (2)


above, i.e. corresponding to the mean squared error
for the th data stream. Further, let the total transmit power be
constrained as .
Let the SVD of be given as , where de- Theorem 1: The IOT criterion based optimal MIMO precoder
note the singular values. The result below now gives the op- which minimizes the sum MSE is given as the solution to the
timal MIMO ZFT for cognitive radio scenarios, which mini- optimization problem,
mizes the net MSE, which is more practical compared to the
sum-rate maximization approaches employed in works such as
[12], [8].
Lemma 1: The optimal ZFT precoder which minimizes the
sum MSE subject to the power constraint i.e. the solution to where . The optimal IOT precoder is ob-
the optimization problem tained as , where,

(3)
1558 IEEE SIGNAL PROCESSING LETTERS, VOL. 22, NO. 10, OCTOBER 2015

where denotes the Lagrange multiplier and is the th diag- Thus, where is the individual MSE
onal entry of the matrix . of the th decoded stream associated with the effective multi-
Proof: Since the objective function above is Schur con- cast group matrix . Let its eigenvalue de-
cave, the optimal precoder is given as . Sub- composition be given as , with denoting its
stituting this in the interference constraint in (2) above yields eigenvalues and . The optimal precoder which mini-
which is equivalent to the constraint mizes the sum MSE bound for this multicast scenario is given
. The optimization problem for the IOT based by the result below, and is signicantly different in nature from
precoder, which minimizes the sum MSE, can be recast as, the work in papers such as [12], [8] which consider only mul-
tiuser performance optimization.
Lemma 2: The optimal ZFT precoder which minimizes the
sum MSE bound i.e. the solution of the optimization problem

It can now be readily seen that the solution of the optimization


problem above is obtained by choosing as given in (3), with
the Lagrange multiplier chosen such that the constraint is sat-
ised.
Further, the optimal IOT precoders which minimize the
product and max MSE can now be obtained employing a
procedure similar to the one described for the ZFT precoder is given as , where , the th entry of is,
design above. With respect to complexity of implementation,
the ZFT has a complexity of to nd the null space
for the projection followed by for the secondary user
SVD where , while the
IOT interestingly only has a complexity of . This Proof: The proof now follows similar to the optimal pre-
compares favorably with schemes in works such as [8], which coder design for the unicast scenario by observing that the quan-
also employ SVD operations for precoder design. tity for the unicast scenario is now replaced by the effec-
tive channel matrix .
III. OPTIMAL MIMO PRECODER DESIGN FOR MULTICAST Similarly, the IOT based optimal precoder for the mul-
COGNITIVE RADIO SCENARIOS ticast MIMO scenario can now be obtained similar to
We now describe the procedure for optimal precoder design the unicast scenario. Set the effective channel matrix
for multicast MIMO cognitive radio scenarios for both the ZFT . The optimal pre-
and IOT based optimization criteria described above. Similar coder is given as , where is given as,
to the unicast scenario, let denote the precoder for the mul-
ticast secondary user group with users. We begin by con-
sidering the ZFT design scenario. Let the optimal precoder
be given as , similar to the unicast scenario. Also, let
denote the effective channel matrix of user
in this scenario, where is the noise plus interference co-
variance for the th user. Consider now the net MSE for the where is the th diagonal element of . However, the
multicast group given as, sum MSE minimization can potentially lead to disproportionate
secondary user power allocation. Therefore, similar to works
such as [13] and [14], one can consider product and max MSE
minimization towards more equitable performance optimization
in cognitive radio scenarios.
where . We now use the Jensens Further, a bound on the objective function for the product
trace inequality, , where MSE in the multicast scenario can be obtained as follows. The
function is convex and the matrices are symmetric. objective function in this case is given in the equation at the
Hence, a lower bound for the sum MSE above is given as, top of the page, where the equivalence of the objective func-
, which implies tions above is due to the fact that is a monotonic concave
function and the inequality in the last step above follows using
the Jensens inequality. The above relation for can be further
simplied as, using the convexity
GUPTA AND JAGANNATHAM: UNICAST/MULTICAST MIMO COGNITIVE OVERLAY/UNDERLAY NETWORKS 1559

of together with the Jensens trace inequality for matrices.


Hence,

Thus, where denotes the determinant.


Therefore, the effective channel matrix can
now be employed to obtain the optimal ZFT multicast precoder
which minimizes the product MSE also, as ,
where is given as, . We now consider
the ZFT precoder design for minimizing the maximum MSE
in the multicast scenario. Let the objective function for this
optimization problem be given as, Fig. 1. BER for unicast transmission.

It can now be shown that . As de-


scribed in [10], the optimal solution for occurs when all the
diagonal elements of the matrix are equal.
Therefore, the objective function can be transformed as
where the optimal precoders for
the optimization objectives and are related as
, where is a unitary matrix such that it has equal diag-
onal elements. Employing the Jensens trace inequality, it can
be seen that,

Fig. 2. BER for multicast transmission.

dominant modes of the interference channel covariance, while


for the IOT, one can use the covariance
Thus, . It follows that the solution in the constraint in equation (2). Fig. 1 demonstrates the BER
to the optimization problem with the optimization objective performance for the optimal ZFT and IOT precoders based
yields the optimal precoder which minimizes the above bound on the sum, product and min-max MSE criteria, together with
on the maximum MSE of the multicast users. Therefore, the op- suboptimal equal, proportional (to ) power allocation and
timal precoder for this scenario is given as , CR-MMSE-BD, CR-ZF-BD schemes from [11]. It can be
, and the matrix is chosen similar to readily seen that the min-max MSE based precoder design
the unicast scenario. We now present simulation results to il- yields a signicant performance improvement in comparison
lustrate the performance of the above precoder designs for uni- to the sum and product MSE designs and the suboptimal
cast/ multicast cognitive radio scenarios. Further, in terms of equal/proportional power allocation based precoders. Further,
computational complexity, it is very interesting to note that the it can be seen that the IOT outperforms the ZFT at larger
overall complexity of precoding for the multicast scenario is values of the interference threshold since tolerating a marginal
interference at the primary users allows the IOT to allocate
, and does not increase with the number of users. power to its strongest modes. Also, while the CR-ZF-BD from
This is possible due to the novel MSE bounds for the various [11] has a poor performance, the CR-MMSE-BD has a slightly
criteria derived in this section. improved performance. However, since CR-MMSE-BD has
a signicantly high computational complexity owing to the
IV. SIMULATION RESULTS
fact that one has to iteratively solve a matrix inversion based
We consider a cognitive radio scenario with primary equation to determine the optimal parameter , our schemes
users, number of transmit antennas at the secondary are more appealing in terms of computational complexity for
base station and antennas at each of the practical implementation. A similar trend is observed for the
primary and secondary users with . Similar to works such multicast scenario with secondary users in Fig. 2.
as [8], [12], [11], we assume the knowledge of the interference In conclusion, this letter presents novel ZFT and IOT pre-
channel which can be achieved through cooperation between coder designs for cognitive radio scenarios. The proposed
the secondary transmitters and primary user. However, in prac- framework is comprehensive and considers diverse criteria
tice, partial CSI regarding the interference channel is sufcient such as sum, product and max MSE minimization. Further, the
for transceiver design in cognitive radio scenarios. For ZFT, various precoder designs have also been extended to multicast
this can be achieved by zeroing the interference along the MIMO cognitive radio scenarios.
1560 IEEE SIGNAL PROCESSING LETTERS, VOL. 22, NO. 10, OCTOBER 2015

REFERENCES [8] R. Zhang and Y.-C. Liang, Exploiting multi-antennas for oppor-
[1] J. Mitola and G. Q. Maguire, Cognitive radios: Making software ra- tunistic spectrum sharing in cognitive radio networks, IEEE J. Sel.
dios more personal, IEEE Pers. Commun., vol. 6, no. 4, pp. 1318, Topics Signal Process., vol. 2, no. 1, pp. 88102, Feb. 2008.
Aug. 1999. [9] K. Cumanan, R. Zhang, and S. Lambotharan, A new design paradigm
[2] S. Haykin, Cognitive radio: Brain-empowered wireless communica- for MIMO cognitive radio with primary user rate constraint, IEEE
tions, IEEE J. Sel. Areas Commun., vol. 23, no. 2, pp. 201220, Feb. Commun. Lett., vol. 16, no. 5, pp. 706709, May 2012.
2005. [10] D. P. Palomar and Y. Jiang, MIMO transceiver design via majoriza-
[3] M. H. Islam, Y.-C. Liang, and A. T. Hoang, Joint power control and tion theory, Found. Trends Commun. Inf. Theory, vol. 3, no. 45, pp.
beamforming for cognitive radio networks, IEEE Trans. Wireless 331551, 2006.
Commun., vol. 7, no. 7, pp. 24152419, Jul. 2008. [11] K.-J. Lee and I. Lee, MMSE based block diagonalization for cognitive
[4] L. Zhang, Y.-C. Liang, and Y. Xin, Joint beamforming and power radio MIMO broadcast channels, IEEE Trans. Wireless Commun., vol.
control formultiple access channels in cognitive radio networks, IEEE 10, no. 10, pp. 31393144, Oct. 2011.
J. Sel. Areas Commun., vol. 26, no. 1, pp. 3851, Jan. 2008. [12] G. Bansal, J. Hossain, and V. Bhargava, Optimal and suboptimal
[5] K. Cumanan, R. Krishna, V. Sharma, and S. Lambotharan, Robust power allocation schemes for OFDM-based cognitive radio systems,
interference control techniques for multiuser cognitive radios using IEEE Trans. Wireless Commun., vol. 7, no. 11, pp. 47104718, Nov.
worst-case performance optimization, in Proc. 42nd Asilomar Conf. 2008.
Signals, Syst., Comput., Oct. 2008, pp. 378382. [13] L. Zheng and C. W. Tan, Optimal algorithms in wireless utility max-
[6] Q. Spencer and A. L. Swindlehurst, Zero-forcing methods for down- imization: Proportional fairness decomposition and nonlinear perron-
link spatial multiplexing in multiuser MIMO channels, IEEE Trans. frobenius theory framework, IEEE Trans. Wireless Commun., vol. 13,
Signal Process., vol. 52, no. 2, pp. 461471, Feb. 2004. no. 4, pp. 20862095, Apr. 2014.
[7] C. Peel, Q. Spencer, A. L. Swindlehurst, and B. Hochwald, Downlink [14] D. Bertsimas, V. F. Farias, and N. Trichakis, The price of fairness,
transmit beamforming in multi-user MIMO systems, in Proc. Sensor Oper. Res., vol. 59, no. 1, pp. 1731, Jan. 2011.
Array and Mult. Signal Process. Workshop, 2004, pp. 4351.

Das könnte Ihnen auch gefallen