Signal Processing - March 2017 - 50

ECDF

with fd denoting the maximum Doppler
quality indicator (CQI) and the feedback
CSi feedback delay is
algorithms described in [7]. More specififrequency shift in hertz, x r = x/Ts being
a significant issue for
cally, the CQI is utilized by the receiver to
the relative feedback delay in multiples of
high-mobility users,
signal via a dedicated feedback link to the
TTIs, and c representing the speed of light.
and it can be a strongly
transmitter which MCS should be employed
The feedback delay has to be seen in relalimiting factor for the rate tion to the temporal variability of the chanfor transmission to achieve reliable as well
as efficient transmission; this information
performance and reliability nel; if the channel is quasi static (as in
is derived from the SINR currently expemany indoor scenarios), a feedback delay is
of wireless transmission
rienced by the receiver. We observe from
irrelevant. It only matters if the channel
in vehicular scenarios.
Figure 3 that, under ideal circumstances of
changes significantly within the duration
delayless noncausal feedback (delay 0 ms),
of the feedback delay, which is gauged by
high-mobility users achieve almost the same performance as
the normalized feedback delay. More specifically, the Dopthose at low mobility. There is only a small performance degpler shift fd is inversely proportional to the coherence time of
radation, which is caused by the need to apply a slightly more
the channel; hence, the product xfd is proportional to the
conservative transmission rate adaptation, since the channel of
number of coherence intervals elapsing during the feedback
high-mobility users varies even within one transmission time
delay duration. The exact proportionality constant depends on
interval (TTI). This observation confirms the expected mobilthe observed Doppler spectrum [8]. In our aforementioned
ity enhancement achieved by dual connectivity. Yet, as soon as
example, we have for x = 1 ms with v = 5 km/h and
we consider certain delay in the CQI feedback link, we observe
v = 150 km/h, x n . 0.009 , and x n . 0.27 , respectively. We
strong throughput degradation at high mobility, with more than
will encounter the normalized feedback delay again later in
25% of subscribers obtaining zero throughput even with only
the section "Future Enhancements and Challenges."
1 ms feedback delay. This performance degradation occurs
CSI feedback delay is a significant issue for high-mobility
because transmission rate adaptation with feedback delay is
users, and it can be a strongly limiting factor for the rate perbased on outdated channel state information (CSI) leading to
formance and reliability of wireless transmission in vehicular
signal outages due to mismatch between the utilized transmisscenarios. The problem is, of course, not specific to dual consion rate and the rate supported by the channel. This loss in
nectivity, but occurs whenever rate adaptation is based on outthroughput goes hand in hand with an increase in latency, since
dated CSI. In the section "CSI Feedback Enhancement," we
each lost packet must be retransmitted. Notice, the TTI length
discuss possible signal processing approaches to mitigate this
of LTE is Ts = 1 ms. Hence, a delay of the feedback processing
performance degradation.
below this value is infeasible; even the value of 5 ms considered in Figure 3 may be hard to achieve in practice.
MBSFN-based V2V communication
To gauge the expected impact of feedback delay x in milWhen active road safety support is implemented in C-ITS,
liseconds, with respect to the system parameters carrier frecommon information must be delivered to many vehicles
quency fc , TTI length Ts, and user velocity o, we define the
within certain geographic regions; e.g., status information,
such as position, velocity and direction, of one vehicle is
normalized feedback delay as
shared with all other vehicles in the vicinity via CAMs to
v
enhance mutual awareness of road traffic participants. Prior
(1)
x n = xfd = x r Ts fd = x r Ts fc ,
c
studies have shown that both a direct exchange of periodically
generated CAMs using DSRC as well as an indirect distribution over dedicated roadside infrastructure, requires a substantial amount of bandwidth to guarantee timely packet delivery. In
1
fact, with a growing number of vehicles within the geographic
0.9
region of interest, the currently foreseen 5.9-GHz band for road
0.8
safety critical communication can easily be overloaded by
Delay 0 ms (0 TTI)
0.7
CAM distribution [9].
0.6
In such situations, mobile cellular networks can be helpful
Delay 1 ms (1 TTI)
0.5
to offload some traffic from a dedicated ITS infrastructure.
0.4
Delay 5 ms (5 TTI)
Employing eMBMS/PWS features of LTE, the broadcast nature
0.3
of cellular systems can be utilized to efficiently deliver the
0.2
5 km/h
same message to many users within certain geographic regions
0.1
150 km/h
formed by MBSFN areas. To demonstrate the capabilities of
0
such an approach, we conduct system-level simulations, compar0
2
4
6
8
10
12
14
ing message distribution via unicast and multicast transmission
Throughput (Mb/s)
[10]. Both approaches have certain advantages/disadvantages:
■ Unicasting in LTE enables per-user scheduling on favorable
Figure 3. The throughput in dependence of user velocity and CSI feedtime/frequency resources to utilize channel and multiuser
back delay in a dense HetNet supporting dual connectivity.
50

IEEE SIgnal ProcESSIng MagazInE

March 2017

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