Cognitive Cooperative Vehicular Ad Hoc Networks ... 2 The State Key Laboratory ofTntegrated Services Networks, Xidian University, China ... Institute, China United Network Communications Corporation Limited, Beijing, 100048, P.R.China.
VEHICULAR NETWORKING
Cooperative Spectrum Allocation with QoS Support in Cognitive Cooperative Vehicular Ad Hoc Networks ZHANG Lep,2,\ LUO Tao!,2, LIU Wei4, ZHU Siting\ LI Jianfeng! 1
Beijing University of P osts and Telecommunications, Beijing, 100876, P.R.China. The State Key Laboratory ofTntegrated Services Networks, Xidian University, China 3 Network Technology Research Institute, China United Network Communications Corporation Limited, Beijing, 100048, P.R.China. 4The 20th Research Institute of CET C, Xi'an, 710086, China 2
Abstract:
To solve the contradiction between
the increasing demand of diverse vehicular wireless applications and the shortage
enhance the communication reliability of safety service considerably in CC-VANET. Keywords:
VANET; spectrum sensing;
of spectrum resource, a novel cognitive
resource allocation; asymmetric; generalized
cooperative vehicular ad-hoc network (CC
nash bargaining solution
VANET) framework is proposed in this paper. Firstly, we develop an adaptive cognitive spectrum sensing (ACSS) mechanism which can help to trigger and adjust the spectrum
49
I. IN TRODUCTION
The vehicular ad-hoc network (VANET)
sensing window according to network traffic
is a comprehensive network, which can ex
load status and communication quality. And
change the safety and non-safety information
then, Generalized Nash Bargaining Solution
among vehicles or between vehicles and fixed
(GNBS), which can achieve a good tradeoff
road side unit (RSU). It can support various
between efficiency and weighted fairness, is
intelligent transportation system (ITS) ap
proposed to formulate the asymmetric inter
plications over the wireless links. In 1999,
cell resource allocation. Finally, GNBS
Federal Communication Commission (FCC)
Safety (GNBS-S) scheme is developed to
assigned 75MHz spectrum at 5. 9GHz for
enhance the Quality of Service (QoS) of
vehicular communication. However, recent
safety applications, especially in the heavy
studies indicate that the channel bandwidth
load status, where the bandwidth demanded
might be inadequate to support the severe dif
and supplied cannot be matched well.
ferent quality of service (QoS) requirements
Furthermore, the primary user activity (PUA)
of safety-service in peak hours of traffic [1-3].
which can cause rate loss to secondary users,
Thus, with the increasing communication de
is also considered to alleviate its influence to
mand of diverse applications and severe QoS
fairness. Simulation results indicate that the
requirements, the limited and static spectrum
proposed CC-VANET scheme can greatly
resource can result in severe message access
improve the spectrum efficiency and reduce
delay and packet loss rate.
the transmission delay and packet loss rate
Recently, cognitive radio (CR) has been
on the heavy contention status. And GNBS
considered as an effective technique to im
spectrum allocation scheme outperforms both
prove the efficiency of spectrum utilization by
the Max-min and Max-rate schemes, and can
allowing cognitive users to share the wireless China Communications· October 2014
channel with primary users in an opportunis
The remainder of this paper is organized
tic manner. And cooperative communication
as follows: related works are firstly reviewed
I n t h i s p a p er, w e
which can increase link capacity by exploiting
in Section 2. Then, Section 3 describes the
propose a novel cen
spatial diversity has also attracted a lot of
proposed CC- VANET architecture, and the
tralized CC-VANET
attention in recent years. For this purpose,
ACSS scheme is developed in section 4.
framework to solve the contradiction be
we propose a novel cognitive cooperative
After that, the basic theory for the proposed
VANET (CC-VANET) to solve the contradic
bargaining solutions and a two-band partition
d e m a n d of d i v e r s e
tion between the increasing demand of diverse
algorithm for inter-cell spectrum allocation are
vehicular services and
vehicular services and the spectrum scarcity
given in section 5. Finally, the performance
the shortage of spec
of vehicular communication in this paper.
evaluation results are described in section 6
trum resource.
The CC-VANET system enables VANET to
and conclusions are drawn in section 7.
tween the increasing
use additional spectrum opportunities outside the standard 5.9GHz band, and it can greatly enhance the QoS performance of various ITS
II. RELATED WORKS
applications. Besides, through cooperative
In order to extend the channel capacity,
spectrum sharing strategies between
researchers attempt to identify spectrum holes
heterogeneous cognitive networks as in liter
of TV band on the road with the character of
ature [4], the primary users might offer some
predictable vehicle mobility in literature [2].
spectrum access opportunities to cognitive
And a fuzzy VANET system is developed in
users, and both the whole system efficiency
[5] to dynamically assign additional spectrum
and fairness can be improved greatly.
from ISM band to a control channel. How
Though cognitive radio and cooperative
ever, it has never considered when and how
communication have made great achievements
to trigger the cognitive mechanism and the
respectively, the research of their applications
contradiction between the cognitive gains of
in CC-VANET is still at a preliminary stage.
additional spectrum and the overhead of spec
Due to the dynamic spectrum sharing of CR,
trum sensing. In this paper, we will address
resource allocation in CC-VANET becomes
the issues mentioned above and develop a
more complicated than the traditional vehicular
novel ACSS mechanism. In [3], a cooperative
networks. The notable primary user activity,
sensing framework with a stationary BS
asymmetric spectrum requirements between
which provides coordination instructions
multi-cells and severe QoS requirements of
to the passing vehicles is proposed. The
safety service should be considered in CC
literature [6] points out the problem with a
VANET. In addition, when and how to trigger
centralized fusion center, and proposes a belief
the cognitive spectrum sensing mechanism
propagation spectrum sensing method which
and the relationship between cognitive
requires each vehicle to send its respective
overhead and performance gains are also of
belief of the presence of primary user. To
great importance to be deeply researched.
guarantee the QoS of vehicular applications,
Considering these factors, we define a
the authors in [7] develop a novel spectrum
communication load metric to evaluate the
management framework including shared
network traffic status, and develop an adaptive
channel, exclusive-use channel and cluster
cognitive spectrum sensing (ACSS) scheme
size control, with the theory of constrained
to trigger and adjust the sensing window
Markov decision process (CMDP). Howev
according to the network traffic load status.
er, the achievements mentioned above have
Then, Generalized Nash Bargaining Solution
never addressed the asymmetric spectrum
(GNBS), which can achieve a good tradeoff
requirements between multi-cells in vehicular
between efficiency and weighted fairness, is
environments. For this purpose, we will model
proposed to formulate the asymmetric inter
the asymmetric resource allocation problem
cell resource allocation.
with GNBS to favor the hot cells in severe
China Communications· October 2014
50
contention status. Game theory which includes cooperative and
the road side unit (RSU) based on CR (CR
non-cooperative methods is widely regarded
RSU) and local information processing unit
as an efficient theory to analyze the resource
(LIPU). In general, we suppose the whole
allocation problem in wireless networks. The
network can be divided into different cognitive
non-cooperative game theory, in which each
subsystems each with several cognitive cells
player is self-interested and the Nash equilib
(CCs). As shown in Fig.l, the system works
rium is often inefficient, has been studied in
as follows: Firstly, the CRVs take the task of
[8] for multi-cell OFDM A resource allocation.
local load estimation and spectrum sensing,
With the aim to minimize the users' transmis
and then periodically report related results
sion power under the constraints of minimal
to the corresponding CR-RSU. Then, the
rate and maximal power, it is used to allocate
CR-RSU is responsible for data fusion and
the subchannels and power to the uplink users.
resource allocation in a single cognitive cell.
The cooperative game approach which empha
Finally, LIPU will calculate and predict the
sizes collective rationality and fairness is also
network load metric of the subsystem and
used to model the spectrum sharing strategy in
then decide when and how to trigger the
[9, 10]. In [9], Nash Bargaining Solution (NBS)
ACSS mechanism. In addition, LIPU is also
is studied in the resource allocation scenario of
in charge of inter-cell resource allocation with
power, rate, and subchannels for a single-cell
cooperative game theory and maintaining the
OFDMA system, and it can obtain more fair
cognitive spectrum pool.
and efficient performance than traditional
As in Fig. 2 , a c o g n i t i v e s u b s y s t e m
allocation algorithms. In order to favor some
composed o f four rectangle cognitive cells is
players, different bargaining powers for users
considered. Assume the density of cognitive
are introduced to NBS for interference channel
vehicles and the bandwidth requirements of
in [10]. In this paper, GNBS is introduced to model
each cell can be obtained by the LIPU. As sume a serious road traffic jam has happened
the inter-cell spectrum allocation problem. For
in cognitive cell-l (CC-l) and it will result in
asymmetric cognitive cells with different load
heavy communication load and QoS degrada
levels and bandwidth requirements, GBNS
tion. However, there may be residual spectrum
scheme can achieve a good tradeoff between
in other adjacent cognitive cells. Thus, the
the weighted fairness and overall achiev
asymmetric traffic statuses and bandwidth
able rate. And then the scheme is extended
requirements among multi-cells should be
to the spectrum starvation case in which the
addressed in CC-VANET. Suppose the avail
QoS support of safety applications should be
able subchannels can be obtained accurately
given high priority. Afterwards, a novel two
by cooperative spectrum sensing technique.
band partition allocation scheme is developed
In addition, the cells in a cognitive subsystem
for different communication load status.
cannot select the same subchannels because
Furthermore, the primary user activity (PUA)
of co-channel interference. The channel state
is also considered to alleviate the degradation
information for each vehicle over different
of fairness.
channels is assumed to be perfectly estimated.
III. PROPOSED SYSTEM 3.1 The proposed CC-VANET
51
tities, i.e., the vehicle based on CR (CRV ),
3.2 System model
In this paper, a two-step scheme is developed for the spectrum resource allocation in CC VANET. Firstly, LIPU is responsible for
architecture
allocating subchannels to each cell fairly
In this paper, we propose a centralized CC
and efficiently, which is named inter-cell
VANET architecture composed of three en-
allocation. Then, RSU will allocate the availChina Communications· October 2014
able subchannels among different CRV s in each CC, which is called intra-cell allocation. Assume each CRY can access the licensed spectrum in an overlay manner. we mainly focus on the asymmetric inter-cell spectrum allocation problem in CC-VANET. Consider a subsystem composed of N CCs, and Nn on-board units (OBUs) are randomly distributed in each cell. Suppose there are
K
available subchannels with equal bandwidth
WHz in the cognitive spectrum pool. The total
achievable rate for a cell can be expressed as
R"
=
I��l a:�kr:
(I)
r: respectively represent the
where a: and
llSL
n m
n m
RS[
Data Pusion Tntra-ce 11 A 11oCiltion
Data fusion Tntril-ce11 Allocation Bruadcast
llepOrT
toopcrclLivc spccLrur
channel indication, and the average utility of the k-th subchannel allocated to the CCn, and /;k denotes a discount factor of the achievable rate. It is assumed that no any subchannel can support transmission for more than one cell, �N Vk, with the definition i.e., L...J " �1 a�
a�
=
{I; ;
=
I,
channel 0th
0
.
k is assigned to CCn,
erwlse.
(2)
The discount factor of the achievable rate /;k=l-Pk, where Pk denotes the probability that the k-th subchannel is reoccupied by the primary users. The average utility of the k-th subchannel is defined as
r:
r;
=
=
(ljN,,)
I,:>:;
Wlog2 (I +
n
E {I, 2, ... ,N}
Fig.1 Framework of the proposed CC-VANET
(3a)
(p;h;)jcr); i E {I, 2, ... ,Nn}
(3b)
r; is the achievable transmission rate of CRY i, h; and p; represent the subchannel gain and transmit power for CRY i in the k-th
where
Licensed User
(3) �f�; � � [":CC----1 � �
....-. .,. ....,.-�
:
-----
11 :
·----
�.-- j � --b.l i .CC �
!
A: �-� �--- A -�-----L ----
-�1 :; I'� :- - - �- -�- - A�- -�T-- - - � I
....
:
;
�
-- -- �
... I
t. fl
� -
----
:
� .
•
:
....
-t��-��- � �----- . "� . I.
:
I
i i
..
----
.....
: -F
1
�
--
fl.:
!1 I
---
.
: I
-
� ---�--T------L �� i -
4.'. -�/ I
-
______
:
I
_
i
Fig.2 Illustrative scenario of a su bsystem in CC-VANET
subchannel, respectively. It is averaged by Nn because there are Nn OBUs. The noise power for all subchannels is assumed to be the same, and equals to
(j2.
we define a communication load metric CJt) to represent the network contention level of the subsystem. As described in (4), the packet
IV. ADAPTIVE COGNITIVE SPECTRUM
loss rate and message delay are considered as
SEN SIN G SCHEME
the most important QoS parameters to eval
In order to guarantee the QoS of safety service,
The message delay can be considered as the
China Communications· October 2014
uate the performance of vehicular services. sum of queuing delay, contention delay due 52
to other vehicles and the transmission delay. Each CRY will compare its obtained packet loss rate
DJt) and message delay PJt) with
the corresponding service-related thresholds.
The basic theories of GNBS and its
If the estimated packet loss rate or message
application in asymmetric spectrum resource
delay exceeds the threshold, the load level will
allocation are given in this section. And we
be added 1. Thus, it is obvious that the local
also propose a simplified two-band partition
load metric for each CRY can only be one
allocation algorithm, and generalize it to the
integer, i.e., C,,(t)
E {O, 1, 2}. C,,(t) ST" . [(D"(t) > D:h) + (p"(t) > P;h)] +(1 - ST,,) . [(D"(t) > D�D + (P n(t) > P�:) ] (4) =
Where STn represents the message type that CRY is transmitting. And its definition as
{0, non
following ST n
=
1,
- safety related service
safety related service
(5)
applications. 5.1 General nash bargaining
solution
As mentioned above, LIPU is considered as inter-cell spectrum allocation, and the CR RSU is supposed as the bargaining player of
information from the CRV s within its communication range, the communication load status for a cognitive cell will be obtained by averaging the measurements using (6). The RSUs send the information to LIPU periodically, and then LIPU will calculate CUPu(t) which represents the average network
contention level of the subsystem. NOJiU C�BU(t) CHSU(t) _1 _ NOllU n
heavy load status to enhance the QoS of safety
the central controller which is in charge of
When an RSU receives the contention
Vn E N }
each cell. Define a nonempty bounded set
{R" E S IR" � R�'i",
, which represents
the set of feasible utilities that each player can obtain if they cooperate after satisfying its minimum utility.R"ll"
=
(R7"ll, ... R�"ll) is
considered as the disagreement point (e.g., minimum rate requirement in this paper).
And f(S, Rmin ) is the outcome of an N-person
bargaining problem and NBS can provide a
L =l
(6)
L C:SU(t)
(7)
come can be solved via optimization problem
We can infer that CRSU(t) and CUPu(t) must
to further generalize the NBS to allow each
=
CLlPU(t)
=
_1_
I'iRSU
NHSU 11=1
be any number between 0 and 2 because
C�BU(t) E {O, I, 2}. Obviously, the larger it is, the more additional spectrum is needed. Con sidering the contradiction between the gains of cognitive spectrum and the overhead of spectrum sensing, we proposed a novel ACSS scheme as in (8).
Ws (t)
=
(W;J2) . C1PU (t)
(8)
W, (t) represents the spectrum sensing window and Wp is the whole bandwidth of where
cognitive spectrum pool. Obviously, the proposed ACSS mechanism can adjust the sensing window according to network traffic load and communication quality.
53
v. COOPERATIVE SPECTRUM
ALLOCATION STRATEGY
unique and fair Pareto optimal operation point under the axioms in [9]. To ensure fairness and efficiency, the out in Nash bargaining model. And it is worthy CC to have different bargaining powers. The GBNS can be expressed as
f(S, Rmi" )
=
where (01,
arg
• • •
RET-�'lli' n�=, (R n - R�i")""
,0,,)
0
E [ , U"
L�=l 0"
=
(9)
1, and
en represents the bargaining power for CCn. Without loss of generality, en can be propor tional to the communication load level or the minimum requirements. And it is defined as equation (10). On �
(Rmin_s + RlllillJlS) (R::u,,-, + R"UllJ
0::
'" a;
:c
�
fairness between multi-cells. JF!
=
(Rj
+
R2)2 /(2R� + 2R;)
(20)
R" represents the average rate for each CRY, and R" RJN,,, nE{I,;2}, JFIE[0.5,1]. where
8
=
It can be observed that the proposed scheme
6
can achieve strict fairness for each CRY under different bargaining powers. However, the
4
differences between the cells in a subsystem
2
cause communication outage for the hot cells
are not considered in Max-min scheme, it may with terrible road traffic conditions. o
2
Consider a scenario that a serious traffic
3
jam or an accident has happened in CC-l. Set
Fig.7 Achieva ble rate for cognitive cells with dife f rent inter-cell allocation schemes
Nj=200, N]=80, there are 160 and 30 CRYs which are requesting for safety applications respectively in CC-l and CC-2. The number of available subchannels in the cognitive
x1� 14ri---,---,---,,---,---,---,----,---,---,---,
spectrum pool ranges from 50 to 80, while they are not enough to support the whole safety applications. And then the outage prob ability for safety applications is evaluated in Fig.) 0. As GNBS-S scheme gives high priority
"' 1 0 :;:,
to the safety applications, it outperforms the
(J)
traditional allocation schemes and can reduce
e ro
0: (J) :0 '" > (J) :E ()
«
the outage probability to a great extent. VII. CON CLUSION
4
In this paper, we propose a novel centralized
2
CC-VANET framework to solve the contra 0.1
0.2
0.3
0.6 0.5 0.4 Bargainng power of CC-1
0.7
0.8
0.9
diction between the increasing demand of diverse vehicular services and the shortage of spectrum resource. Firstly, the ACCS mechanism is developed to adjust the sensing
Fig.8 Achieva ble rate with GNBS allocation scheme
81 can be seen in an intuitive sense. [t is shown
that the cognitive cell with larger bargaining power can obtain more spectrum resource using GNBS scheme. And it can achieve com parable fairness with Max-min scheme when the requirements of both cells are similar. The results demonstrate that the generalized bargaining schemes can allocate the resource 57
window according to communication traffic load level and link quality. And then a cooperative bargaining spectrum allocation based on GNBS is proposed to formulate the inter-cell resource allocation in CC VANET. With the introduction of bargaining power, the generalized scheme considering PUA can allocate the spectrum resource in a weighted fair manner and favor the hot cells with severe traffic jams or emergency
China Communications· October 2014
- -- - -- - - -- - -- ,
incidences. In addition, an improved scheme
1.1 "
with QoS support for safety service is also developed especially for the heavy load allocation algorithm is proposed for the Simulation results demonstrate that the ACSS mechanism can offer additional spectrum for vehicular communication and reduce the message delay and packet rate loss greatly. And the proposed GNBS scheme can improve the communication reliability considerably
,--
1�·
cases. Furthermore, a practical two-partition two cell scenario in different load statuses.
--;-
>