![](https://csdnimg.cn/release/download_crawler_static/15603101/bg4.jpg)
3948 Ouyang et al.: Greedy Relay Selection in Large-scale Wireless Networks
blocked for reasons (e.g. fading, security), which means the successful transmission can only
be achieved with the assistance from relays. Owing to this,
should select relays that satisfy
its demand before the procedure of transmission. For the sake of simplicity, it is assumed that
the user can only select and utilize at most one relay during the transmission.
The network is decentralized so that no base station coexists for the information collection
or broadcast. Due to the scale of the network, nor do the relays exchange their local
information with each other, and therefore each individual relay in the proposed network only
possesses the knowledge of its own. In addition, since the spectrum resource is scarce in a
large-scale network, we assume that all the activities related to one user are limited in one
single channel. Under these conditions, it is impossible for every relay to upload its real-time
status through the “feedback” channels to the user for comparison.
From above observations, we model the proposed network as a “blind” network for two
reasons: (1) for the user, it is blind to all the relays information since it has no access to the
relay numbers, nor to their CSI or positions; (2) for the relays, as is previously mentioned, they
are blind to each other.
We adopt the selective decode-and-forward (SDF) [1] protocol in transmission so that the
relay only forwards the successfully decoded data to the destination. The channel between any
two nodes is assumed to be independent and identically distributed (i.i.d.) with path loss and
Rayleigh fading. The channel gain is modeled as
, where
denotes the distance
between the communicating nodes, the path loss factor is 2, and
is an exponential distributed
random variable with unit variance, respectively. Since most of the real world wireless
networks are mobile, we further assume that the source and the relays can dynamically change
their locations between fixed-length frames to promote the practicality of the model. The
frames are assumed to be time-slotted and each frame contains
slots. We also assume that
each node is equipped with a positioning module (e.g. GPS) for its real time positions. Under
above conditions, a relay serves in one frame may fail to provide even the basic service in next
frame if it moves to a remote location or simply refuses to cooperate. As a consequence of this,
the mobility of the network requires that periodical selection should be performed in each
frame to guarantee reliable transmission.
3. Slotted ALOHA Based Relay Selections
According to the proposed blind scenario and demand of the user, it is known that each frame
consists of a selection phase to obtain a relay and a transmission phase to transmit the data with
the selected relay. To attain an acceptable time-averaging transmission rate in each frame, the
applied relay selection scheme should consider both the instantaneous rate of the relay and the
slots consumed in selection. Before the definition and formulation of the SAGRS scheme, first
we introduce two related slotted ALOHA based relay selection schemes, namely, the “best”
relay selection and the “first” relay selection.
3.0.1 “best” relay selection
The slotted ALOHA based “best” relay selection is the most conventional and yet simple
random access selection scheme. As presented in Fig. 2(a), the length of the selection phase is
fixed in all frames. Its basic concept is: after the broadcast & identification slot (same as the
SAGRS scheme, which will be developed later), the available relays randomly access the
channel in each slot of the selection phase with slotted ALOHA protocol to compete for the
selection. If a relay accesses successfully, it will be stored for comparison. At the end of the