Signal Processing - March 2017 - 51

diversity. Furthermore, it supports dynamic link adaptation
(multiantenna precoding and transmission rate), as well as
selective PHY repetition of lost packets using hybrid automatic repeat request (HARQ). These features facilitate efficient and reliable data transmission to single users with
low retransmission latency of lost packets.
■ Multicasting in LTE is restricted to employ certain
reserved subframes (multicast subframes) and, hence, cannot freely exploit channel diversity. In general, channel and
multiuser diversity in multicasting are more difficult to harvest since many users are served in parallel, and, therefore,
chances of finding time/frequency resources that are favorable for all users simultaneously are small. Nevertheless,
scheduling/resource allocation gains are possible by avoiding the worst time/frequency resources of all users, such
that signal outages become less likely. Multicast transmission in LTE does not support PHY retransmission of lost
packets and, thus, has to rely on slow (in terms of latency)
higher-layer protocols. Furthermore, transmission rate
adaptation must be performed according to the channel
quality of the worst users to guarantee reliable delivery to
all multicast users, reducing the spectral efficiency of the
system; indeed, currently existing proposals do not utilize rate adaptation at all but rather rely on fixed rate
transmission to avoid CSI feedback from the users.
Finally, multicasting in LTE also does not support multiantenna transmission, even though multicast-specific
beamforming/precoding and space-time coding has the
potential to substantially improve efficiency and reliability of data transmission.
In the following, we investigate the performance of
CAM distribution in cellular networks serving vehicles
that move at v = 150 km/h. We consider fixed-rate transmission (employing CQIs 3, 6, and 9), since rate adaptation
in multicasting is not yet supported by the LTE standard.
We conduct system-level simulations for an MBSFN area
consisting of three macro base stations embedded within
a larger cellular network. Within the MBSFN area, CAMs
of size 300 bytes are exchanged among 21 vehicles (seven
per base station) with a periodicity of 100 ms. Hence, for
multicast transmission within the MBSFN area, the total
traffic load is 300·8 bit/user·21 users/100 ms . 0.5 Mbit/s,
since each packet is synchronously broadcast from all
three base stations. In case of unicast transmission, on the
other hand, each vehicle individually receives the CAMs
generated by the other 20 vehicles. Since seven vehicles are
attached to each base station of the MBSFN area, this implies
a unicast traffic load of 300·8 bit/user·20 users·7/100 ms .
3.4 Mbit/s. We determine the overhead for the cellular network caused by CAM distribution, as well as the operationally critical parameters latency and message loss probability
for supporting active road safety in C-ITS. We assume transmission with extended cyclic prefix for unicast and multicast operation.
The results of the simulation are summarized in Table 1.
We observe that a unicast transmission with CQI 3 is unable

to support the generated data traffic, i.e., the network overhead caused by CAM distribution is equal to 270% of the
capacity (assuming 5-MHz bandwidth); thus, the message
loss probability is very high (2 63% ) since many CAMs
have to be dropped. With multicasting, however, the overhead is reduced to 60% and the message loss probability
is below 1%. Yet, even with multicasting in our simulation
scenario, it is not possible to sustain the generated CAM
network load with CQI less than three, because the transmission efficiency would be too low. Notice that, at most, six
subframes per radio frame (consisting of ten subframes) can
be reserved for MBSFN operation in LTE. A very important
metric for active road safety is latency. Since eMBMS does
not support retransmission of lost packets, latency accumulates in multiples of the message generation period (100 ms)
in case of packet loss. Thus, even short signal outages can
severely increase latency; correspondingly, latency of multicasting deteriorates with increasing transmission rate. For
active road safety, latency below 100 ms has to be achieved
[11]. Notice that Table 1 only presents radio-link latency for
downlink transmissions, neither accounting for uplink from
vehicles to base stations nor for CAM distribution within
the MBSFN area; hence, we present the percentage of users
with downlink latency below 50 ms to incorporate a safety
margin with respect to the prescribed 100 ms. We observe
in Table 1 that LTE is not able to sustain this latency with
sufficiently high probability in our simulation. Hence, performance improvements of multicasting in LTE are required
to enable dependable support of road safety applications.
In the section "Multicast Enhancements," we discuss
potential enhancements of LTE to improve multicast transmission, employing dynamic link adaptation and coordination of multipoint transmission. In addition to enhancing
the wireless transmission, we also see potential in optimizing the C-ITS protocol itself. As mentioned previously,

Table 1. A comparison of unicast and multicast transmission
for CAM distribution in cellular networks.
Rate

Metric

Unicast

Multicast

CQI3

Overhead

270%

60%

(0.377 bit/sym)

Message loss prob.

> 63%

< 1%

Latency < 50 ms

n/a

93%

Latency < 250 ms

n/a

98%

CQI6

Overhead

86%

20%

(1.176 bit/sym)

Message loss prob.

Latency < 50 ms

9.5%

80%

Latency < 250 ms

85%

93%

CQI9

Overhead

42%

10%

(2.406 bit/sym)

Message loss prob.

11%

Latency < 50 ms

49%

54%

Latency < 250 ms

74%

76%

IEEE SIgnal ProcESSIng MagazInE

March 2017

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