Signal Processing - March 2017 - 57

max

min SINR u

f ! C N tot # 1 u ! U MBSFN

(9)

subject to:
Sb f

2
2

# Pb, 6b ! B MBSFN

Leak ur # L ur , 6ur ! Ur MBSFN,
with f denoting the joint beamformer applied over all
N tot = / b ! B MBSFN N t, b transmit antennas N t, b of the set of
B MBSFN base stations within the MBSFN area. The parameter
SINR u represents the SINR of user u from the set U MBSFN of
users served via multicasting in the MBSFN area and Leak ur
denotes the interference leakage caused to user ur from the set
Ur MBSFN of users outside the MBSFN area; it is upper bounded
by the leakage constraint L ur . Finally, S b represents a selector
matrix, which chooses the entries of f corresponding to the
transmit antennas of base station b, and Pb is the power constraint of base station b . This problem is nonconvex but,
similar to [20], is amenable to SDR. The operating point of
this multicast interference network with spatially distributed
multicast transmitters can be optimized by coordinating the
interference-leakage constraints of interfering base stations.
Such coordination can be achieved by applying the dual
gradient optimization proposed in [21].
Notice that the discussed beamforming approaches are all
based on explicit CSIT, and, hence, the CSI quantization problematic discussed in the section "CSI Feedback Enhancement"
applies to these methods as well. Furthermore, since the methods are iterative in nature, they exhibit intrinsic delay to achieve
convergence. In vehicular environments, it might be necessary
to employ suboptimal heuristics that are derived from optimal
schemes; such heuristics are also provided in [21]. Also, especially for road safety applications, outage probability and latency
minimization might be of higher importance than rate maximization. Optimization problems for such targets, however, are hard
to formulate due to potential nonstationarity of wireless channels
in vehicular environments [23]. Still, for limited time frames,
assuming quasi-stationary conditions may be valid to formulate
outage-constrained optimization problems.

Scheduling and resource allocation
A major weakness of eMBMS in the LTE standard is its
inflexibility of multicast transmission; i.e., a fixed amount of
subframes within each radio frame is reserved for multicasting. One reason for this is that multicast transmissions within
MBSFN areas suffer from a larger delay spread due to singlefrequency transmission of the same signal from multiple spatially distributed base stations. To compensate for the
increased delay spread, multicast subframes employ the
extended cyclic prefix (16.7ns ) of LTE, whereas unicast subframes mostly use the shorter normal cyclic prefix ( 4.7ns ) to
minimize waste of bandwidth. According to our work on link
adaptation, however, utilizing the extended cyclic prefix may
not pay off, even if intersymbol interference (ISI) occurs; in
many cases of realistic SNR, it is more efficient to simply

accept the remaining ISI and use a more robust transmission
rate to compensate for it [24]. Hence, the LTE standard should
be extended to support multicasting in MBSFN areas with a
short cyclic prefix. This then also allows the incorporation of
multicast transmissions into normal subframes, facilitating a
better dynamic resource assignment according to traffic
requirements, as well as the exploitation of channel diversity
in multicasting. Within the 3GPP, such extensions are considered in the study item on single-cell point-to-multipoint transmission; within this study, single-cell multicasting over the
downlink shared channel is evaluated, which until now has
only handled unicast data transmissions.
A big challenge from a signal processing perspective is efficient scheduling and coordination of multicast transmissions.
In the context of vehicular communications, this involves
dynamic formation of MBSFN areas, to limit the amount of
CAMs that must be distributed while still providing mutual
awareness over sufficiently large geographic areas. Furthermore, it implies determining groups of multicasting users that
must share CAMs due to close spatial proximity. The multicast
scheduler should also be able to decide to offload users with
very poor channel quality to unicast transmissions, to support
selective PHY retransmissions via HARQ, such as to reduce
latency and to improve reliability. First results in this direction are provided in [25], where the authors consider optimizing proportional fairness of concurrent multicast and unicast
transmissions. Heterogeneity of user channel conditions is considered in [25] by optimally partitioning multicast users into
groups, so that users with good signal strength do not suffer
by being grouped together with users of poor signal strength.
A further challenge comes up when the system supports direct
V2V transmission in addition to cellular-assisted transmission;
then the scheduler must further group vehicles according to
direct and assisted transmission.

Further research topics to enhance wireless
vehicular communications
Channel estimation at high mobility
At very high mobility, accurate CSIR estimation using pilot
signals can become challenging due to strong temporal
variation of the wireless channel. Prior investigations have
shown that LTE does not achieve MIMO spatial multiplexing
gains at high mobility due to insufficient density of pilot symbols in the time domain [26]. This low density of pilot symbols can cause poor accuracy of channel estimation and, as a
consequence, unreliable symbol detection. A remedy can be
provided by adapting the pilot pattern to the Doppler- and
delay-spread of the channel. These two properties characterize the coherence time and bandwidth of the channel, respectively, i.e., the time/frequency intervals over which the
channel stays approximately constant. Such an adaptive pilot
scheme can efficiently be realized with minimal feedback
information about coherence time and bandwidth from the
users, requiring an update whenever channel statistics vary
significantly [26].

IEEE SIgnal ProcESSIng MagazInE

March 2017

Table of Contents for the Digital Edition of Signal Processing - March 2017