Mobile Netw Appl (2015) 20:713–724 715
resource blocks that are used by the cellular users in the
proximity. Zhang et al. [10] propose a graph-based resource
allocation method for cellular networks with underlay D2D
communications. In [11], a new interference cancellation
scheme is designed based on the location of users. Yu et al.
[12] propose to use Han-Kobayashi rate splitting techniques
to improve the throughput of D2D communications. To
improve the capacity of cellular networks, the authors of
[13] propose a joint D2D communication and network cod-
ing scheme. The authors of [14] propose an algorithm for
power allocation and mode selection in D2D communi-
cation underlaying cellular networks. The authors of [15]
aim to minimize the overall transmission power in a multi-
cell OFDM cellular network. Different from the works on
underlay mode, the authors of [16–18] propose to allocate
dedicated resources for D2D communications.
Another D2D mode is Out-Band which D2D commu-
nications occur on an unlicensed spectrum such as ISM
2.4G that is not overlapping with the cellular spectrum.
Out-Band D2D is advantageous because there is no inter-
ference issue between D2D and cellular communications.
Otherwise, the user equipments (UEs) operated on the unli-
censed spectrum require more radio interfaces to support
the Out-Band D2D communication, such as WiFi-Direct
[19], Bluetooth [20] and ZigBee [21], and this will increase
the UE hardware design complexity. There are also sev-
eral literatures that refer to Out-Band D2D communications.
Zhou et al. propose to use ISM for D2D communica-
tions in LTE. They propose to group D2D users and allow
only one user per group to contend for the WiFi channel
because simultaneous channel contention from both D2D
and WLAN users can dramatically reduce the network per-
formance [22]. The authors of [23–25] propose to form
clusters among cellular users who are in range for WiFi
communication and only the cluster member with the high-
est cellular channel quality communicates with the BS. The
authors propose a CH selected method based on the Chan-
nel Quality Indicator (CQI). As the algorithm described,
each CNM sends the CQI message to the CH, the CH col-
lects the CQI information and sends it to the eNB. After
that the CNM with the highest CQI is selected to be the
CH by the eNB. In the case of stable channel condition,
the disadvantage of this method is to make the CH on work
for a long time, then the CH’s charge will be consumed
faster than the CNMs, this will lead the result of short clus-
ters lifetime. Furthermore, the work in [26] elaborates on
the required modification of LTE and WiFi technologies
and defines a protocol stack to connect both of them. This
paper sheds light on the different aspects of integrating LTE
and WiFi such as channel quality feedback, scheduling and
security.
3 D2D system model
In this section we study the hybrid D2D architecture and
cluster formation via the Bluetooth interface. As illustrated
in Fig. 1, the scenario includes a BS and two clusters. Inside
the cluster all the UEs transfer traffic to each other directly
via the WiFi-Direct interface. Each cluster has its own
cluster head (CH) which carries out the D2D communica-
tions in the In-Band mode. We call this hybrid architecture
as the Hybrid-Band.
3.1 Out-Band D2D based on Bluetooth and WiFi-Direct
(i) Cluster Formation
Today’s mobile devices are usually driven by the
limited battery power. This makes it essential to
devise an effective scheme that reduces power con-
sumption. In our proposal a low-power Bluetooth
interface is used as the primary option to make clus-
ter formation. In this paper, the terms such as cluster,
CH and cluster normal member (CNM) are defined in
[26].
The mobile UEs that we consider in this paper
are popular user terminals. We assume a UE which
is able to measure its residual charge and signal
to interference plus noise ratio (SINR) has a cel-
lular network interface, a WiFi-Direct interface and
a Bluetooth interface. The cluster has 1 CH which
is capable of calculating all of the CNMs’ Intent
Value ( IV)as(5) defines, and 7 CNMs which are
time-synchronized with the CH via the Bluetooth
interface.
As Fig. 2 shows, the procedure describes how the
cluster is formed. When a mobile UE intends to ini-
tiate a D2D communication, it detects the message
broadcasted
periodically by the cluster. If the mobile
UE detects the message within the maximum time
interval T , it means that there exists a cluster nearby.
Otherwise, there is not any cluster in the available
range. Then the mobile UE initiate to form the cluster
itself.
When there is a cluster around, a new user who
intends to make a D2D communication has to open
its Bluetooth interface firstly. Then it starts to detect
and analyze the CH’s broadcast messages. Secondly,
the mobile UE sends the requested message to the CH
for joining to the cluster. After receiving the request,
the CH makes a response to the mobile UE. As Fig. 3
illustrates, if the mobile UE is permitted to join the
cluster, it will send the confirmed message to the clus-
ter, finally the CH will add the new joining UE’s