3056 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 61, NO. 12, JUNE 15, 2013
Robust MIMO Precoding for Several Classes of
Channel Uncertainty
Jiaheng Wang, Member, IEEE, Mats Bengtsson, Senior Member, IEEE,BjörnOtt
ersten, Fellow, IEEE,and
Daniel P. Palomar, Fellow, IEEE
Abstract—The full potential of multi-input multi-output
(MIMO) communication systems relies on exploiting channel state
information at the tra
nsmitter (CSIT), which is, however, often
subject to some uncertainty. In this paper, following the worst-case
robust philosophy, we consider a robust MIMO precoding design
with deterministic
imperfect CSIT, formulated as a maximin
problem, to maximize the worst-case received signal-to-noise ratio
or minimize the worst-case error probability. Given different types
of imperfect C SIT
in practice, a unified framework is lacking in
the literature to tackle various channel uncertainty. In this paper,
we address this open problem by considering several classes of
uncertainty
sets that include most deterministic imperfect CSIT
as special cases. We show that, for general convex uncertainty
sets, the robust precoder, as the solution to the maximin problem,
can be effic
iently computed by solving a single convex optimiza-
tion problem. Furthermore, when it comes to unitarily-invariant
convex uncertainty sets, we prove the optimality of a channel-di-
agonali
zing structure and simplify the complex-matrix problem
to a real-vector power allocation problem, which is then analyt-
ically solved in a waterfilling manner. Finally, for uncertainty
sets d
efined by a generic matrix norm, called the Schatten norm,
we provide a fully closed-form solution to the robust precoding
design, based on which the robustness of beamforming and uni-
for
m-power transmission is investigated.
Inde
xTerms—Convex uncertainty set, imperfect CSIT, max-
imin, MIMO, minimax, saddle point, Schatten norm, unitarily-in-
variant uncertainty set, worst-case robustness.
Manuscript received July 17, 2012; revised January 5, 2013; accepted March
10, 2013. Date of publication April 12, 2013; date of current version May
20, 2013. The associate editor coordinating the review of this manuscript
and approving it for publication was Prof. Y.-W.Peter Hong. This work was
supported in part by the National Basic Research Program of China (973) under
2013CB336600 and 2013CB329204, National Natural Science Foundation of
China under 61201174, Natural Science Foundation of Jiangsu, China under
BK2012325, and by the European Research Council under the European
Community’s Seventh Framework Programme (FP7/2007-2013) / ERC grant
agreement no. 228044, and by the Hong Kong RGC 617911 research grant,
and by the National Key Special Program of China no. 2012ZX03003005-003.
J. Wang was with the Signal Processing Laboratory, ACCESS Linnaeus
Center, KTH Royal Institute of Technology, SE-100 44 S toc kh o lm, Sweden.
He is now with the National Mobile Com munications Research La boratory,
Southeast University, Nanjing, China (e-mail: jhwang@seu.edu.cn).
M. Bengtsson is with A CCESS L inn aeu s Center, Signal Processing Lab-
oratory, KTH Royal Institute of Technology, S to ckholm SE- 10 0 44, Sweden
(e-mail: mats.bengtsson@ee.kth.se).
B. Ottersten is with ACCESS Linnae us Center, Signal Processing Laboratory,
KTH Ro y a l Institute of Te chnology, Stockholm SE-1 00 44, Sweden , and a lso
with th e Interdisciplinary Centre for Secur ity, Reliability and Trust (SnT) , Uni-
versity of Luxembourg, Luxembourg-Kirchberg L-1359, Luxembourg (email:
bjorn.ottersten@ee.kth.se; bjorn.ottersten@uni.lu).
D. P. Palomar is with the Hong Kong University of Science and Technology,
Hong Kong, and also with the Centre Te cnológic de Te lecom municacions de
Catalunya-Hong Kon g (CTTC- HK), Hong Kong (e-mail: paloma r@ust.hk).
Color v ersions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org.
Dig
ital Object Identifier 10.1109/TSP.2013.2258016
I. INTRODUCTION
I
T is well known that the performance of multi-in
put multi-
output (MIMO) comm unication system s depends
,toasub-
stantial extent, on the quality of the channe
l stat e information
(CSI) [1]. The full benefits of MIMO channels
are achieved by
exploiting CSI at the transm itt er (CSIT) an
d adopting proper
precoding techniques [2], [3]. With p
erfect CSIT, the optimal
MIMO precoding has been well studied
under various criteria
[4]–[6] for either single-user or mu
lti-user com munications. In
practice, however, CSIT is seldom p
erfect but subject to some
uncertainty due to many practical
issues, such as i naccurate
channel estimation, quantizati
on o f CSI, erroneous or outdated
feedback, and time delays or fre
quency offsets between the re-
ciprocal channels. Therefo
re, the imperfection of CSIT has to
be considered in MIMO preco
ding designs so that the commu-
nication system, on one han
d, can fully utilize CSIT, and on the
other hand, is robust to va
rious i m perfect CSIT.
In the literatur e , imperfe
ct CSI is modeled by either stochastic
or de terministic approac
hes. The stochastic model assumes that
the channel is a random q
uantity and its instantaneous value
is unknown but its stat
istics, such as the mean and/or the co-
variance, is known by
the transmitter. In this case, the robust
design usually aims a
t optimizing either the long-term average
performance [7]–[1
0] or the outage performance [11]–[14]. On
the o ther hand, the
deterministic model, which is mo re suit-
able to characteri
ze instantaneous CSI with errors, assumes that
the actual chann
el lies in the neighborhood, often called the
uncertainty se
t or region, of a nominal channel known by the
transmitter. T
he size of this set represents the amount of un-
certainty on t
he channel, i.e., the larger the set is the more un-
certainty th
ere is. In this case, a precoding design is said to
be ro bust if i
t can achieve the best performance in the worst
channel wi
thin the uncertai nty set, or equ ival entl y can guar-
antee a per
formance level for any channel in the uncertainty set.
Such robu
st precoding desig ns can be achieved by optimizing
the worst
-case performance [15], often leading to a maximin or
minimax
problem [16]–[32]. Note that, worst-case robustness is
related
to statistical robustness in some situat ion . For example,
many ou
tage based robust designs are often transformed into d e-
termi
nistic formulations by defining an uncertainty set that has
a cert
ain probability [11], [12], [14].
The fo
cus of this paper is on worst-case robust precoding de-
signs
based on deterministic imperfect CSIT. As an important
bran
ch of robust designs [15], the philosophy of worst-case
rob
ustness has been widely used in signal processing [16]–[19]
1053-587X/$31.00 © 2013 IEEE