1
ROPAS: Cross-layer Cognitive Architecture for Mobile UWB Networks Chittabrata Ghosh, Student Member, IEEE, Bin Xie, Senior Member, IEEE and Dharma P Agrawal, Fellow, IEEE
Abstract—The allocation of bandwidth to unlicensed
users while limiting the average power transmitted in each
users, without significantly increasing the interference on
sub-band. In our proposed novel ROPAS architecture,
the existing licensed users, is a challenge for Ultra Wide-
dynamic channel allocation is achieved by a CR-based
band (UWB) networks. Our research work presents a novel
cross-layer design between the PHY and Medium Access
Rake Optimization and Power Aware Scheduling (ROPAS)
Control (MAC) layers. Additionally, the maximum number
architecture for UWB networks. Since UWB communica-
of parallel transmissions within a frame interval is formu-
tion is rich in multipath effects, a Rake receiver is used
lated as an optimization problem. This optimal decision is
for path diversity. Our idea of developing an optimized
based on the distance parameter between a transmitter-
Rake receiver in our ROPAS architecture stems from the
receiver pair, bit error rate and frequency of request
intention of reducing the computation complexity in terms
by a particular application. Moreover, the optimization
of the number of multiplications and additions needed for
problem improvises a differentiation technique among the
the weight derivation attached to each finger of the Rake
requesting applications by incorporating priority levels
receiver. Our proposed work uses the Cognitive Radio
among user applications. This provides fairness and higher
(CR) for dynamic channel allocation among the requesting
throughput among services with varying power constraint and data rates required for a UWB network.
C. Ghosh, B. Xie, and D. P. Agrawal are with the Department of Computer Science, University of Cincinnati, Cincinnati, OH-45221 USA. e-mail:
[email protected], {xieb, dpa}@cs.uc.edu This work is supported by National Science Foundation (NSF) under Grant No. NeTS-WN0721641.
Index Terms—Cognitive Radio, Joint Power and Frequency Allocation, Power Aware Scheduling, Primary and Secondary Users, Ultra Wideband.
2
I. I NTRODUCTION
U
be significantly improved with dynamic channel sensing and allocation of licensed bands, performed
LTRA Wideband (UWB) [1] techniques by a specially-designed radio technology called are used extensively for higher data rates Cognitive Radio (CR) [2]. CR is a revolutionary
over short transmission ranges, especially for multechnology that substantially improves spectrum eftimedia traffic. UWB spans over a 7.5 GHz (3.1ficiency with the aid of advanced spectrum sensing 10.6 GHz) bandwidth at a power level near the and dynamic channel assignment in licensed bands, noise floor. However, the UWB transmissions can but without actually obtaining a license. CR is increase the noise floor and introduce interference primarily a software-defined radio that is aware on adjacent channels. This may result in substantial of its surrounding environment [3], and is capable performance degradation to the existing users. To of sensing spectrum used by neighboring devices, avoid such undesirable situations, Federal Commuchanging frequencies, adjusting output power, or nications Commission (FCC) defined the spectral even altering parameters and characteristics of its mask of -41.3 dBm/MHz for the UWB transmistransmission. sions in certain overlapping bands. For example, the On the other hand, UWB communications are radar and satellite systems span over a bandwidth of rich in multipath effects. In this paper, we make 1.6 GHz (3.1 - 4.7 GHz). Therefore, power control a novel effort in utilizing Rake filtering in our promechanisms and reduction of channel interference posed Rake Optimization and Power aware Schedulare important design issues in a UWB receiver. ing (ROPAS) architecture to exploit effects of the Given limited radio spectrum available to acmultipath channels. In our ROPAS design, the commodate an increasing number of users, the unique multi-objective Rake optimization ensures underlying system must adapt to the exponentially minimal bit error rate (BER) for the receiving growing demand of spectrum access. Static chanapplications while strategically selecting an optimal nel allocation of licensed bands to SUs results in number of multipaths among the many possible wastage of channel bandwidth, and in turn, lowers propagation paths. The joint optimization helps in spectrum utilization. The spectrum utilization can
3
reducing the computation complexity at the Rake “free” (i.e., idle) channel detection, interference receiver while minimizing interference for better
measurement, multipath selection, and power con-
estimation of transmitted bits.
trol. For example, the interference and channel
The design of the medium access control (MAC) fading characteristics have to be considered from the layer for the UWB network is significantly different PHY layer so that the data frames can be transmitted from the traditional multihop ad hoc networks. The
with less interference. Furthermore, transmission
reason for this difference in design is due to the
power control over the links is performed by the
transmission power restrictions imposed on such
MAC layer with an objective of limiting power
UWB signals in the licensed bands by the FCC.
while satisfying the packet transmission requirement
The MAC layer protocol design should possess the
(e.g., delay, distance). Thus, the cross-layer sharing
salient features: (i) minimize power consumption
of information between these two layers is indis-
due to communications to reduce interference on
pensable, especially for CR-based mobile nodes in a
the licensed users and (ii) fair and efficient sharing
dynamic network, since the channel conditions vary
of resources between communicating devices.
randomly with varying mobility patterns. At the
Now, power aware MAC layer with minimum
same time, in ROPAS, the cross layer based power
interference can be efficiently developed when in-
control and link scheduling strategy helps the CR
formation can be exchanged with the PHY layer
in imposing a limit on the transmit power in each
before allocating links to data traffic. This exactly frame interval. The CR uses increased transmission is the central idea of our novel ROPAS architecture.
power for delay sensitive traffic to reach the desti-
In this paper, we take advantage of cross-layer
nation in minimum number of hops and therefore,
design between the PHY and MAC layers for CR
minimum delays while it employs reduced transmis-
aided dynamic channel assignment in the UWB
sion power for delay tolerant services.
network. However, the CR-based dynamic channel
In our cross layer design and simulation, we have
assignment can not be achieved in a straightforward chosen the UWB (3.1-10.6 GHz) for link allocation manner, and involves inter-related issues such as
and for traffic communication purposes due to its
4
high data rate on each sub-channel of 528 MHz
II. R ELATED W ORK
bandwidth [1]. The main contributions of this paper Dynamic channel allocation in mobile networks are as follows: has been addressed by previous works in myriad ways. The distributed, fault-tolerant allocation [4] •
We present a multi-objective optimization problem for the Rake receiver to reduce computational complexity,
•
We present the cross-layer multi-objective optimization design for dynamic frequency selection with optimal transmission powers allocated to each subframe, an optimal partition of a licensed sub-band, and
•
We discuss the cross-layer design of the CRbased priority based scheduling while supporting maximum number of parallel transmissions within a frame interval.
depends much on the channel usage information of the interfering neighbors of a requesting user, based on which it can compute the best channel allocation. This strategy involves a high message complexity due to exchange of channel usage information of all its one-hop neighbors. The other research on dynamic channel allocation [5] is based on a mutual exclusion between the “request” and “reject” messages for allocating a group of channels. But, this strategy suffers from the problem of fairness among requesting users with additional problem of message complexity due to exchange of “request” and “reject” messages. An improved version of the
The rest of this paper is organized as follows. channel allocation is discussed in [6] where channel Section II deals with the previous work on dynamic
allocation is based on the ratio of the serviced
channel allocation strategies. Our proposed algo-
data rate to the required data rate. Additionally,
rithm, ROPAS is described in Section III. Section
the power distribution algorithm in this research
IV illustrates the optimal power allocation strategy work “adjusts” the average transmit power of the based on distance and priority differentiations. Sim-
channel based on the received signal-to-noise ratio
ulation results are discussed in Section V followed per user. But this allocation strategy may give rise by the conclusion in Section VI.
to a serious problem: assume that a channel is
5
allocated to a loss-sensitive application and the
location based MAC layer protocol and routing
channel suffers from deep fading while their power [8] depends on the distance information to achieve distribution strategy decides to decrease (“adjust”)
low power levels and increased lifetime and im-
the average transmit power on this channel. This
proved network performance. Saving of power in
will result in loss of the entire transmission due to
each UWB device is achieved by using a sleeping
poor channel conditions and cause a dramatic per- interval when the devices are not participating in formance degradation. Our proposed ROPAS cross-
active transmissions. Again this paper works on
layer architecture has a solution to this problem.
the trade-off between power constraints and net-
Scheduled links are allocated power by the CR
work latency when our main focus is for power
based on the retrieved information about channel
constrained protocol design. The application of CR
conditions from the physical layer.
[9] deals with the network layer perspective in
The design of efficient MAC layer protocol with
UWB networks. The routing decisions are made
low power consumption and strict computational
on the basis of a cognitive cost function that takes
complexity is a persisting research challenge. The
care of important issues like synchronization, end-
authors in [7] have dealt with these issues using the
to-end delay and coexistence with licensed users.
margin based power allocation scheme to maintain
Our proposed architecture deals with a cross-layer
power constraint along with an exclusive region MAC optimization protocol where the CR in each based scheduling algorithm to reduce interference.
UWB device decides upon allocating channels to its
But this research work suffers from the inherent
applications within the imposed power constraints.
problem of centralized decisions taken by a cen-
Additionally, proportional fairness is achieved by
tralized controller for power allocation and slot
the unique priority scheduling performed by the CR.
allocation. Therefore, the system becomes much
III. T HE ROPAS A RCHITECTURE
more complex with the increase in number of UWB The central idea of cross-layer design is to imdevices, resulting in heavy overhead due to sharing prove the overall performance of wireless networks of varying users with the centralized controller. The with the exchange of information between the differ-
6
tween the PHY and MAC functionalities as shown
1
MAC Layer
3
Cognitive Radio CR Manager Manager (Priority Scheduling)
PowerChannel aware Scheduler Scanner
in Fig. 1. All the central decisions are taken by the
6
Channel Power aware Scanner Scheduler
CR Manager (CRM) which also has the capability
4
7
7
Physical layer Channel Estimation
5 Rake Optimization
5
of interaction between different modules. The other 2
Interference Measurement
PHY Layer
two modules in the MAC layer are responsible for efficient dynamic channel allocation among mobile nodes with limiting power constraints. The Channel
Fig. 1.
Cross-layer design of the ROPAS architecture
Scanner (CS) divides the entire UWB into smaller sub-bands and scans these sub-bands in periodic
ent layers of the protocol stack. This was not possiintervals for possible “free” (not used by licensed ble in traditional wireless networks. Recent research users) channels. The next module is called the on cross-layer design [10]-[11] showed a substantial Power-aware Scheduler (PAS) which aims at a improvement in the routing efficiency, throughput, multi-objective joint power control and link schedulfairness and delay variance among different appliing of data frames. Additionally, it also performs the cations. Physical layer information exchange [12] hybrid queuing strategy to achieve fairness among with the MAC and network layer has also exhibited requesting applications. The three modules at the superior network performance. We made an effort bottom of Fig. 1 are associated with the PHY layer in utilizing multi-objective optimization techniques of each node in the network. One of the modin the CR-based cross layer design which involves ules is the Interference Measurement (IM) which the MAC and PHY layers. measures the interfering power sensed in each subThe ROPAS architecture is shown in Fig. 1 and band due to users in adjacent sub-bands. The PAS the entire protocol is described in several steps (i.e. works with the IM to limit transmission power in marked by the numbers). In the proposed ROPAS, any particular sub-band within the permissible limits in addition to traditional CR modules in the PHY (- 41.3 dBm/MHz [1] or 0.039 mW/528 MHz for and MAC layers, several functional modules are UWB communications). The Rake Optimization included in order to improve the collaboration be-
7
Module (ROM) deals with the PHY layer Rake re-
•
Step 2: The CS with ready reference to the
ceiver. This multi-objective optimization computes
IM module for probable interference power,
a minimal number of Rakes or fingers needed by
detects the “free” channels. It is noted that the
the Rake receiver for maximum signal power and
IM module is located in the PHY layer.
hence maximum signal-to-noise ratio (SNR) at min-
•
Step 3: Upon the response from the IM, the
imal BER. The Channel Estimation Block (CEB)
CS sends the detected free channels with their
estimates the fading condition of the channel as
respective identifications to the CRM. There-
well as the channel error rates. The CEB shares the
fore, the CRM has now the complete informa-
cross-layer information with the CRM to select the
tion about the free channels for the requesting
best link (in terms of fading and error rate) for data
applications.
transmission among the adjacent one hop neighbors.
•
Step 4: The CRM requests the PAS module for
Our proposed architecture addresses these two
transmit power limits on each of these “free”
issues: the dynamic channel allocation for the trans-
channels. The CRM also sends information
mitting applications and the Rake optimization for
about the delay constraints for the requesting
receiving processes. The Rake optimization is a
application.
purely PHY layer issue and utilizes the interference
•
Step 5: The PAS refers to the IM for signal-
power information from the IM to optimize the
to-interference and noise ratio needed for joint
number of propagation paths selected for minimal
power control and link scheduling; the PAS
BER. For dynamic channel allocation, let us con-
module divides the MAC layer frames into
sider an example to get a better understanding of
subframes (based on delay constraints of the
our cross-layer design of ROPAS. Seven steps are
requesting applications) and assigns a group
involved in the entire channel allocation process:
of links to each subframe. The module also allocates a group of transmit powers based on
•
Step 1: An application arrives at the CRM with the delay constraints. its link request and delay constraint. The CRM •
refers to the CS for possible “free” channels.
Step 6: The PAS sends the information to the
8
CRM about the frame interval, fraction of each
A. Rake Optimization Module
subframe and probable group of transmit power In this subsection, we discuss the Rake optimizaallocation to each subframe. tion, which is an enhanced Rake receiver module •
Step 7: Finally, the CRM checks with the CEB in the PHY layer shown in Fig. 1. The Rake for probable error rates and fading conditions optimization employs a multi-objective optimization based on the information received from the strategy for an optimal selection of multipaths out PAS module. Then, the CRM allocates the of the several possible propagation paths. power constrained links to the frames deterA Rake receiver is generally used for UWB mined by the PAS. networks. Since UWB communications are rich in multipath effects, Rake receivers are used to accumulate the energy in significant multipath components. It consists of a bank of correlators or
It can be seen from these steps that for a power fingers where each finger is synchronized to a constrained link an appropriate collaboration be-
multipath component. The output of each finger is
tween different modules enables the data trans-
coherently combined using different techniques like
mission in an optimal manner in terms of current
Maximal Ratio Combining (MRC) [13], Minimum
channel and link status (i.e., delay constraint of
Mean Square Error, etc.
the requesting user, utilization of the channel with
The complexity in computing the Rake receiver
power constraints, and the interference of link). The
output involves two parts: (i) multiplications of
collaboration considers two critical issues in the
{N × M } matrix with {M × N } matrix gives
subframe transmission: the channel characteristics
O(M N 2 ) [13] and (ii) additions of the above two
for dynamic channel allocation and the transmission
matrices of similar dimensions results in a complex-
power for reducing the interference. In the following ity of O((M − 1)N 2 ), where M and N denote the subsections, we will describe each module in detail
number of correlators and the weights assigned to
and illustrate the interactions between each module.
each correlator respectively. Our idea of developing
9
an optimized Rake receiver stems from the intention
our optimal selection of a few fingers will reduce
of reducing the computation complexity in terms of
M and the corresponding reduction in weights will
the number of multiplications and additions needed
reduce N .
for the weight derivation attached to each finger
Let K = {1, 2, · · · , k, · · · , K} be the set of all
of the Rake receiver. We have chosen MRC Rake the multipaths. The energy-to-noise ratio (EN Rk ) receiver for its lower computation complexity when compared to other Rake receivers.
in k-th multipath can be written as [1]: EN Rk =
To illustrate our assertion, we assume that the i-th
Pk × τ c , N0 × W × σT
(2)
where Pk is the average power received in the kreceived signal at time instant t is ri (t). The output th multipath, τc is the coherence bandwidth of the of the Rake receiver yi (t) for the i-th received signal UWB channel, N0 is the one-sided power spectral with R fingers or correlators can be given by: yi (t) = γ T ×
R−1 X
density of the background Additive White Gaussian ri (t − δj ),
(1)
j=0
Noise (AWGN), W is the signal bandwidth and σT
where γ = [γ0 , · · · , γR−1 ]T are the weights associ-
is the standard deviation of the AWGN noise within
ated with each finger of the Rake receiver, T is the
the symbol duration, T .
transpose operation and δj is the delay associated
The Rake optimization is to strategically select
with j-th correlator of the Rake receiver to capture
only a few of the multipaths out of all the possible
the multipath signal from its predefined delayed
ones. The reason behind this is two fold: (i) received
path. Now, the computation complexity depends on
signal energy from each and every multipath may
the number of fingers used and their corresponding
not improve the total desired signal energy at the
finger weights. In order to reduce the computation
Rake receiver and (ii) delayed multipaths may suffer
complexity, we can strategically select an optimal
from severe fading or may have been corrupted
number of fingers out of many multipaths in UWB
due to channel interference, thereby resulting in
communications. If the value of M and N can be
increased BER. The idea is to optimize the num-
reduced, then the computation complexity can be
ber of multipaths chosen so as to maximize the
reduced to a great extent. Hence, the basic idea of
EN Rk for the k-th path. On the other hand, the
10
optimization needs to minimize the overall system
0
f0 (k) and f1 (k ) can be rewritten as:
BER, which implies minimization of the overall bit energy, Eb . Therefore, it becomes a multi-objective
Pk max f0 (k) = PU −1 i
j=0,j6=i
Pjk
, k = 0, · · · , K − 1,
0
optimization.
0
max f1 (k ) =
KX −1
P l.
(5)
l=0
The multi-objective function in multipath k with power Pik for the i-th UWB receiver in the presence
To solve this multi-objective functions, we can
of interfering U nodes can be represented finally as
either create Pareto-optimal charts [14] and select
below:
the best solution from the same or combine them as Ã
max PU −1
j=0,j6=i
done here. In fact, another approach is to select a
!
Pik Pjk
, k ∈ K.
(3)
set of real values λi which refers to the multiplier for the i-th maximizing objective function, f0 (i).
Then maximizing Eq. (3) for all k ∈ K.
Hence, our new objective function L(φ) becomes:
µ
min
Eb N0
¶
K−1 X
= min
Pik +
k=0
= min
ÃK−1 X
!
Pk
U −1 X
K−1 X
0
0
L(φ) = f1 (k ) −
Pjk
KX −1
λi × f0 (i).
(6)
i=0
j=0,j6=i k=0
over all the users.
This is still a combinatorial optimization problem.
(4)
To reduce it to a linear Integer Programming (IP)
k=0
problem [14], we introduce a set of variables Xi 0
0
Next, let φ = {1, 2, · · · , k , · · · , K } be a selection
defined as:
from K i.e., φ ⊂ K. Hence, our goal is to choose a subset φ that maximizes the power given in the first
Xi = 1, if multipath is selected and,
objective while maintaining a low BER. Therefore,
= 0, if multipath is not selected.
0
K is the optimized number of paths in the set of paths φ, and K is the number of multipaths in the
Therefore, the problem in Eq. (6) can be
set of possible propagation paths. Therefore, the op-
reformulated over the set X (constitutes individual
timization problem with two maximizing functions
Xi0 s) as:
11
P ower[i][j] with j 6= 1 is the received interference power from i-th multipath. We have also declared the multiplier, λi (i.e., λ[i] in Fig. 2) and the binary variable, Xi (i.e., Xi or x[i] in Fig. 2) for our joint optimization. The transmission power constraint on each multipath for UWB communication is limited to 0.039 mW. Therefore, for any correlator, we Fig. 2.
Pseudo code for the Rake multi-objective optimization
impose a power constraint of 0.039 mW as any multipath with higher power values may be corrupted due to interference. With this underlying logic for
0
0
L(X) = X × f1 (k ) − " KX −1 0
=
i=0
KX −1
λi × f0 (i)
i=0
our optimization problem, we solve Eq. (7) for the
# i
Xi × Pj Xi × P l − λi P , i l6=j Pl
subject to Xi ∈ [0, 1].
optimal selection of paths as shown in Fig. 2. (7)
It is easy to see that this is a linear IP problem
B. Interference Measurement
and can be easily solved using a standard solver
The IM is another module in the PHY layer
like the Branch-and-Bound method. We have used
involved in Steps 2 and 5 as shown in Fig. 1.
GLPK [15] (version 4.10) for solving this multi-
This module is needed to calculate the signal-to-
objective optimization problem.
interference noise ratio (SINR) which estimates the
Our implementation of the IP and the pseudo-
ratio of power due to the allocated link to power
code is shown in Fig. 2. First, we declare a variable, due to other adjacent interfering links at a softP ower[i][j] which represents the power received by
decision variable. Let us assume that M information
the Rake receiver from the i-th multipath carrying
bearing symbols, Sk (1), · · · , Sk (M ), independently
information of the j-th user. P ower[i][1] is the
and identically distributed (i.i.d) are chosen from a
power received by the desired UWB Rake receiver finite set with zero mean. The mean E[Sk (m)] and (j = 1) from the i-th multipath. This implies that
variance, E[| Sk (m) |2 ] for the k-th link are defined
12
as:
allocation and the state of the wireless channel. Gk,l ≥ 0, l 6= k is the path gain between the link
E[Sk (m)] = 0, and
l and link k. Therefore, if the transmit power on
E[| Sk (m) |2 ] = Pk ,
link l is Pl , then the expected interference on 1 ≤ m ≤ M, 1 ≤ k ≤ L,
(8)
link k 6= l is Pl Gk,l . Additionally if Gk,l = 0,
where, Pk represents the signal power on the k-th then the link k is said to be orthogonal to link link.
l. Again, Gk,k ≥ 0 represents the self and inter-
Then, the expected value of SINR at the k-th link among L links is given by:
dispersive nature of the wireless channel. σk2 > 0 is
G k × Pk , 2 l=1 Pl × Gk,l + σk Pk
the Gaussian noise variance at the output of link, k.
SIN Rk (P ) = PL = P L
l=1 Pl ×
symbol interference which occurs due to the time
³
Gk,l Gk
´
+
σk2 Gk
C. Channel Estimation Block (CEB)
, (9)
1 ≤ k ≤ L,
where Pk ≥ 0 is the transmitted power on link k. We further define the transmit power vector, P as
The CEB module is involved in Step 7 of our cross-layer dynamic channel allocation strategy. The minimum mean square estimation (MMSE) [16] algorithm runs at the CEB to determine existing chan-
P = (P1 , · · · , PL ) ∈
