Signal Processing - November 2016 - 113

consider (PT1, PR 1), whose transmission is relayed by ST2 during a fraction
PU Transmitters
PU Receivers
b 1, 2 (0 # b 1, 2 # 1) of T, while also
receiving a fraction l 1, 2 (0 # l 1, 2 # 1)
PR1
PT1
of C from ST2 . The parameters l 1, 2
and b 1, 2 are referred to as the price and
PR2
PT2
time-slot allocation numbers, respecSU Transmitters (Relays)
..
...
tively, whose exact values will be deterPRN
.
PTN
ST2 ST1
mined by the matching algorithms.
During the cooperative relaying
stage in the initial slot 1, a fraction x 1, 2
STM
(0 1 x 1, 2 1 1) is first allocated for PT1
SU Receivers
Time Slots for PTl Transmission ST3 . . .
to broadcast its signal to ST2 and PR 1 ,
(Via Relay)
SR2
SR1
occurring in the first b 1, 2 x 1, 2 T time
Time Slots for
slots. In the subsequent slot 2, ST2 coopSlot 1
Slot 2
STq Transmission
SRM
eratively relays the signal from PT1 to
...
PR 1 . PR 1 then applies maximum ratio
PTl 's Money
SR3
STq 's Money
combining to the signal received from
Paid by STq
PT1 in the first b 1, 2 x 1, 2 T time slots
(slot 1), and the signal received from FIGURE 2. The SU and PU spectrum-access model.
ST2 in the subsequent b 1, 2 (1 - x 1, 2) T
time slots (slot 2). After this cooperathe performance metric of the nondecision users, which will
tive relaying stage, PT1 ceases transmission, allowing ST2 to
lead to unacceptably low performance for these users. Here
transmit to SR 2 over the spectrum occupied by (PT1, PR 1) in
matching theory approaches are able to obtain a better perforthe final (1 - b 1, 2) T time slots. It is notable if the amplify-andmance, one that is desirable for the spectrum access strategy to
forward relaying protocol is considered, x 1, 2 = 1/2.
get all users involved in the decision-making process.
To summarize the matching algorithm, each PT will first
In [30] and [7], the proposed distributed spectrum access
make an offer to the ST that is first in its preference list. The
framework is based on the interaction of both sides, both the
ST will then check if the offering PT is in its preference list.
PUs and the SUs. Moreover, most of the work in CR literature
If it is, and the ST is already matched with another PT, the ST
[31]-[33] considers a scenario with only one PU, while matchhas two choices.
ing-based models can cover multiple players from both PUs
1) If the offering PT can provide a better utility than the ST's
and SUs with or without transfer.
current matching, then the ST will reject its current matching in favor of the new matching.
2) If the offering PT cannot provide a better utility than the
Channel allocations in opportunistic spectrum access
ST's current matching, the ST will reject the PT's offer.
There are many works in the literature that discuss opportuIf the ST is not matched, then the ST will be matched with
nistic channel selection in CR networks; however, most of
the offering PT. If the offering PT is not in the ST's preferthem are based on two assumptions. The first assumption is
ence list, the ST will reject the offering PT.
that when a channel is available, all the SUs can have similar
Note that if the ST rejects a PT, then the PT will update its
performance using that channel. The second assumption is
proposal, and the PT will either 1) decrease its price-allocation
that the full knowledge of channel parameters or information
number by a price-step number e or 2) decrease its time-slot
about other users is always available. However, these assumptions are not realistic in all scenarios.
allocation number by a time-slot step number, depending on
Due to the dynamic environment in PUs, spectrum availwhich option maximizes the PT's utility and assuming a posiability may be different for different SUs even in the same
tive price and time-slot allocation number and the minimum
channel. Moreover, available spectrum spaces are time varydata rate requirement for the PT is satisfied. The algorithm will
ing due to the random traffic of PUs. When there are multiple
then repeat this procedure with each PT until no more matchPUs and SUs with different channel information and different
ings are possible.
performance for each channel, modeling such scenarios is conAs discussed previously, a network management mechasidered challenging without using matching theory. In [34], a
nism needs to be incentive compatible, that is, the users are
channel allocation problem without the limiting assumptions is
motivated to participate and accept the results. However, in the
considered, where each SU only considers its individual history
majority of current game theory techniques used in CR, only
and its local observation and independently makes a decision
one type of users is considered as the primary decision makers.
to access a channel. To improve the network's throughput, it is
In other words, either the PUs or the SUs are involved in the
desirable to design a dynamic and distributed spectrum-shargame to determine access to the spectrum resources [29]. In
ing strategy. In [34], by using matching theory, a decentralized
these approaches, the decision makers do not take into account
IEEE Signal Processing Magazine

November 2016

113

Table of Contents for the Digital Edition of Signal Processing - November 2016