arXiv:1408.5204v1 [cs.IT] 22 Aug 2014
Signal and Interference Leakage Minimization in
MIMO Uplink-Downlink Cellular Networks
∗
Tariq Elkourdi,
∗
Osvaldo Simeone,
†
Onur Sahin and
‡
Shlomo Shamai (Shitz)
∗
CWCSPR, New Jersey Institute of Technology, Newark, NJ, 07102, USA
†
InterDigital Inc., Melville, New York, 11747, USA
‡
Department of Electrical Engineering, Technion, Haifa, 32000, Israel
Email: {tariq.elkourdi,osvaldo.simeone}@njit.edu, onur.sahin@interdigital.com, sshlomo@ee.technion.ac.il
Abstract—Linear processing in the spatial domain at the
base stations (BSs) and at the users of MIMO cellular systems
enables the control of both inter-cell and intra-cell interference.
A number of iterative algorithms have been proposed that allow
the BSs and the users to calculate the transmit-side and the
receive-side linear processors in a distributed manner via message
exchange based only on local channel state information. In
this paper, a novel such strategy is proposed that requires the
exchange of unitary matrices between BSs and users. Specifically,
focusing on a general both uplink- and downlink-operated cells,
the design of the linear processors is obtained as the alternating
optimization solution of the problem of minimizing the weighted
sum of the downlink and uplink inter-cell interference powers
and of the signal power leaked in the space orthogonal to the
receive subspaces. Intra-cell interference is handled via minimum
mean square error (MMSE) or the zero-forcing (ZF) precoding
for downlink-operated cells and via joint decoding for the uplink-
operated cells. Numerical results validate the advantages of the
proposed technique with respect to existing similar techniques
that account only for the interference power in the optimization.
Index Terms—Linear precoding, interference alignment, up-
link, downlink, MIMO cellular system.
I. INTRODUCTION
Linear processing in the spatial domain at the base stations
(BSs) and at the users of a Multi-Input Multi-Output (MIMO)
cellular system is a well studied technique that enables the
control of both inter-cell and intra-cell interference (see, e.g.,
[1]). A number of iterative algorithms have been proposed in
the past few years for the design of the linear processors that
are either centralized, see, e.g., [2] and references therein, or
can be instead implemented in a decentralized way [1][3]-[7].
In the latter case, the BSs and the users calculate the transmit-
side and the receive-side linear processors in a distributed
manner via message exchange based only on local channel
state information.
The distributed techniques in [1][3]-[7] differ in the infor-
mation that is exchanged between the BSs and users and in
the processing that is carried out at the two sides. Another key
classification of these techniques can be done with respect
to methods that apply to MIMO interference channels, i.e.,
cellular systems with a single user per cell, and techniques
are suitable for to more general cellular systems with multiple
users per cell. The interference leakage minimization (ILM)
techniques of [3][4] require the exchange of unitary matrices
Downlink Inter-cell interference
Desired signal
Cell 3
Cell 2
Cell 1
Uplink inter-cell interference
Figure 1. Multi-cell uplink-downlink MIMO system. Downlink and uplink
inter-cell interference signal paths are shown for a user in cell 1 as an example.
between the two sides
1
and was proposed for a MIMO interfer-
ence channel. References [6][7] generalize the ILM technique
to a cellular system with an arbitrary number of users per
cell, where the cells operate in either uplink or downlink. In
contrast, the technique proposed in [1] requires the exchange
of additional information beside unitary matrices and applies
to the downlink of a general MIMO cellular system. In this
regard, we observe that the transmission of unitary matrices
is facilitated by the advances in the quantization over the
Grassmann manifold (see, e.g., [9]) and is hence desirable,
making the ILM scheme of [3][4][6][7] potentially more viable
for practical implementation. The signal plus interference
leakage minimization technique (SILM) of [5] modifies the
ILM strategy by including in the cost function, not only the
interference power, but also the power of the signal that is
wasted in the space orthogonal to the receive subspaces. This
scheme also requires the exchange of unitary matrices and was
studied in [5] for MIMO interference channels.
In this paper, a novel iterative strategy is proposed that
generalizes SILM [5] to a MIMO cellular system with an
1
A different implementation based on pilot symbols and estimation is also
possible, see [8].