ZHOU et al.: A COOPERATIVE MATCHING APPROACH FOR RESOURCE MANAGEMENT IN DYNAMIC SPECTRUM ACCESS NETWORKS 1049
in cellular DSA, the problem of dynamic resource supply
with various demanders should be well investigated [20].
From the perspective of quality of service for SUs’ demands,
in [21], Alshamrani et al. proposed a spectrum allocation
framework for heterogeneous SUs in real time and non-real
time (NRT) applications, respectively. In [22], Sodagari et
al. proposed a time-optimized and truthful dynamic spectrum
rental mechanism. In [23], H. Zhou et al. introduced a packing
approach to fast and optimally allocate the time-frequency
blocks. In [24], Yuan et al. discussed a dynamic time-spectrum
blocks allocation problem in cognitive radio networks. In
[25], C. Singh et al. introduced a provider-customer matching
resource allocation strategy based on the coalitional games. In
[26], N. Zhang et al. investigated a maximum weight matching
problem for the cooperative DSA in multi-channel cognitive
radio networks. However, none of these works are specific for
cellular n etworks and consider the aforementioned features of
cellular DSA.
III. N
ETWORK MODEL AND PROBLEM FORMULATION
A. Dynamic Spectrum Access Service Model
We consider a dynamic spectrum access scenario which
consists of a cellular network as a primary network, a local
spectrum sensing network and a secondary network as shown
in Fig. 1. In the cellular network, the base station (BS) man-
ages the resource scheduling to serve mobile stations (MSs)
that are referred to as PUs. Due to different communication re-
quirements of PUs, the statically pre-assigned time-frequency
resource blocks are different. In the meantime, the MSs have
different characteristics of spectrum usage behaviors [6][7],
such as the frequent variations in time and space domain.
Similar to [10], the secondary network is self-organized in the
same area. Once a transport lin k is requested for a realtime
bulk data flow transmission between two SUs, e.g., video
conference, data for warding and multi-m edia service, etc., SUs
will apply to external sensing agents for the DSA opportunities
with appropriately sized bandwidths and spectrum access
durations.
The local spectrum sensing network is composed of the
common sensor nodes and sink sensor nodes. The sink sensor
nodes can obtain the realtime channel prediction information
and provide the dynamic spectrum access opportunities for
SUs. Once PU turns on in the free spectrum bands, the r ealtime
spectrum usage update made by sensor nodes will inform SUs
to stop transmission tasks to avoid the interference to PUs.
Definition 1: (TFBs Supply) The available TFBs sup-
ply set from external sensing agents at time t is de-
fined by RB
t
= {f
t
p
1
,f
t
p
2
, ..., f
t
p
n
}, where the avail-
able TFBs supply function of external sensing agents is
f
t
p
i
(κ
t
p
i
,α
t
p
i
,β
t
p
i
, Δ
t
p
i
,ρ
max,t
p
i
,π
t
p
i
), ∀i =1, 2, ..., n, n is the
available time-frequency block number, κ
t
p
i
is channel band-
width, α
t
p
i
and β
t
p
i
are the arrival time and ending time,
respectively, Δ
t
p
i
is the time-slot size, Δ
t
p
i
= β
t
p
i
−α
t
p
i
, ρ
max,t
p
i
is the permitted transmission power, and π
t
p
i
is the required
price for the TFBs.
Definition 2: (TFBs Demand) For SUs with TFBs de-
mand, the set of n TFBs demanders is defined by Φ=
{μ
1
,μ
2
, ..., μ
n
}, and the user demands set is defined by
TABLE I
S
UMMARY OF IMPORTANT SYMBOLS
Symbol Definition
p
i
Primary user i
μ
j
Secondary user j
f
t
p
i
Resource block supply function of p
i
at time t
N Resource block number provided at time t
κ
t
p
i
Channel bandwidth in f
t
p
i
Δ
t
p
i
Time-slot size in f
t
p
i
ρ
max,t
p
i
Permitted transmission power in f
t
p
i
π
t
p
i
Required price in f
t
p
i
γ
†
µ
j
Resource block demand function of μ
j
at time t
κ
†
µ
j
Required channel bandwidth of μ
j
at time t
Δ
†
µ
j
Applied time-slot of μ
j
at time t
ρ
†
µ
j
Transmission power ability of μ
j
at time t
π
†
µ
j
Accepted leasing price of μ
j
at time t
RB
t
The available time-frequency block supply set at time t
RB
t
p
The massive sized time-frequency block supply set at time t
RB
t
d
The small sized time-frequency block providing set at time t
Θ
t
p
i
Evaluated transmitting data capacity
φ
†
µ
j
The real transmitting data of SU μ
j
N The coalitional player set
v Spectrum sharing payoff function
B Realtime coalition structure
ζ
f
t
µ
j
Real resource consuming cost of SU μ
j
at time t
ζ
c
t
µ
j
Under utilized resource cost shared by SU μ
j
at time t
D
†
= {γ
†
μ
1
,γ
†
μ
2
, ..., γ
†
μ
j
},wherethej-th user γ
†
μ
j
=
{κ
†
μ
j
, Δ
†
μ
j
,ρ
†
μ
j
,π
†
μ
j
}, κ
†
μ
j
is the required channel bandwidth,
Δ
†
μ
j
is the applied time-slot, ρ
†
μ
j
is the power transmission
ability, and π
†
μ
j
is the acceptable leasing price.
In Definition 1, we assume that the sensing function of
cooperative agents can guarantee the short detection time of
vacant spectrum [17][27], and the release time of spectrum
vacancy information t
is no late than the resource available
time α
t
p
i
, i.e., α
t
p
i
− t
≥ 0. Hence, the released spectrum
information can satisfy both the demands of online dynamic
spectrum sharing and offline spectrum reservation. In Defini-
tion 2, all the parameters can be calculated according to the
factual transmitting data volume, required data rate and power
constraint conditions, etc. For the delay-sensitive SUs, th ey
can reserve the realtime TFBs before transmission, to avoid
the problems caused by the channel reservation delay.
Generally speaking, all the external sensing agents act like
the local sellers in a spectrum mar ket [28]-[30] . At tim e
instant t, each sensing agent will publish TFBs information.
There are a random number of supplied TFBs that will be
traded among the n independent SUs with different demands.
We assume that all the realtime information provided by the