Title: Spectrum%20Opportunity-Based%20Control%20Channel%20Assignment%20in%20Cognitive%20Radio%20Networks
1Spectrum Opportunity-Based Control Channel
Assignment in Cognitive Radio Networks
- Loukas Lazos, Sisi Liu and Marwan Krunz
- ECE Dept., University of Arizona, Tucson
- Presented by Loukas Lazos
- SECON 2009, Rome, Italy
2The Promises of Cognitive Radio Technology
- Solve two critical problems
- Spectrum scarcity Exploit dynamic spectrum
opportunities - Interoperability Communicate with their own
kind and other radio technologies - Co-existence with legacy users (Primary radios)
- CRs must obey regulatory rules Higher priority
to Primary Radio (PR) users - Policy enforcement heavily coupled with CR
hardware and protocol design
orthogonal frequency bands
CRA
PRA
CRB
CRC
PRB
3Cooperative Diversity
- Nodes cooperate in the spectrum sensing process
CRC
CRD
CRA
PR1
PR2
CRB
CRE
orthogonal frequency bands
Exchange individual sensing observations to
define idle channels Share idle channels need a
mechanism for negotiation
The existence of a coordination (control) channel
is required!
4Current Practices for Control Channel Assignment
- Use unlicensed bands such as ISM bands
-
Already overcrowded Uncontrolled interference No
guaranteed performance
5Current Practices for Control Channel Assignment
- Allocate a slice of spectrum for carrying control
traffic - Contradicts the open spectrum architecture
Finding an unoccupied frequency band is a
challenge
We need a fixed licensed frequency band to build
dynamic spectrum allocation systems
FCC
Cognitive Radio Advocates
6Dynamic Control Channel Assignment
- Allocate one of the idle channels for control
- Creates the following circular dependency
7Further Challenges
- Spectrum opportunities vary with location and
time - Leads to a partition of the network into clusters
- Need for dynamic migration of the control channel
based on PR activity - Need for inter-cluster coordination
orthogonal frequency bands
8Spectrum-opportunity Based Assignment
- Five-step process
- Sense idle channels
- Discover neighbors (in the absence of a common
channel) - Exchange idle channel list
- Agree on a common time schedule for the
control-channel location - Migrate control channel if a PR user occupies the
current one
CRC
CRD
CRA
PR
CRB
CRE
9Neighbor Discovery
- In the absence of a control channel, CRs may
reside in different frequency bands - Construct a universal time-slotted schedule
- Each CRi i individually determines the list of
idle channels Cii i1,,ik - A CRi i beacons its channel list Ci on channel
ij ? Ci during slots t 1, 2, if ij
(t-1) (mod M)1, and stays silent otherwise (M
number of channels). - Any CRk that hears CRis transmission places
CRi in Nk - CRk i communicates with CRi ? Nk using the
channel schedule derived from Cii until a common
control channel is setup.
10Neighbor Discovery
- Construct a universal time-slotted schedule
t
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
t
CRA
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
t
CRB
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
t
CRC
1 2 3 4 5 6 7 8
9 10 11 12 13 14 15 16
wasted slots 50 efficiency
orthogonal frequency bands (M)
11Time Synchronization Issue
- Time synchronization need not be tight
t
1 2 3 4
5 6 7 8
9
t
CRA
1 2 3 4
5 6 7 8
9
t
CRB
1 2 3 4
5 6 7 8
9
t
CRC
1 2 3 4
5 6 7 8
9
orthogonal frequency bands (M)
12Cluster-based Control Channel Assignment
- Partition the network into clusters
- Take into account local idle channels
13Mapping Clustering to a Graph Problem(1)
- Combine network topology with idle channel
availability
Info available at CR A
CA 1, 2, 3, 4, 5, 6, 10 CE 2, 3, 5, 7
CB 1, 2, 3, 5, 7 CF 2, 4, 5, 6, 7, 10
CC 1, 2, 3, 4, 10 CG 1, 2, 3, 4, 8
CD 1, 2, 3, 5, 7 CH 1, 2, 5, 8
Network connectivity graph
Each CRi constructs a bipartite graph Gi(Ai, Bi,
Ei)
14Mapping Clustering to a Graph Problem(2)
- Compute a biclique Qi(Xi, Yi) (complete
subgraph) of Gi(Ai, Bi, Ei)
A biclique Qi(Xi, Yi) represents a cluster with
membership Xi where channels Yi are common to all
cluster members
15Problem Biclique Construction
- Design Criteria
- Maximum edge biclique problem Maximize the
number of edges in Qi - Provides a balance between cluster size and of
common channels - Known to be NP-Complete Peeters 2003
- Weighted maximum edge biclique problem Maximize
the weighted sum - Takes into account the quality of each channel
- Also NP-Complete Dawande et. al. 2001
- Constrained maximum edge biclique problem
Maximize edges subject to a constraint on the
cardinality of one or both sets Xi, Yi
16Heuristic for Maximum-Edge Biclique Computation
A
D
B
C
G
H
1
2
3
4
5
6
10
Xi x Yi Xi x Yi
7 12
10 15
9 10
CA 1, 2, 3, 4, 5, 6, 10 CE 2, 3, 5, 7
CB 1, 2, 3, 5, 7 CF 2, 4, 5, 6, 7, 10
CC 1, 2, 3, 4, 10 CG 1, 2, 3, 4, 8
CD 1, 2, 3, 5, 7 CH 1, 2, 5, 8
17Distributed Clustering Algorithm
- Spectrum-Opportunity Clustering (SOC)
- CRs individually compute their cluster
memberships by solving the maximum edge biclique
problem (or a variant). - CRs broadcast the computed cluster membership
information to their neighbors, and update
cluster memberships according to a total biqlique
ordering. New cluster information is
rebroadcasted. - CRs compute the final cluster membership
information and broadcast one more time to ensure
consistency - Can show that
- All CRs individually reach to an agreement with
respect to clusters and common idle channels - At least one CR is within one-hop range of all
others in the cluster can serve as clusterhead
(CH)
18Control Channel Migration
CRA
CH
CRB
PR
19Clustering Performance
20Further Problems to be Considered
- Need for re-clustering under various PR activity
models - Required reclustering frequency
- Development of local repair algorithms to avoid
global reclustering - Heterogeneity in channel quality
- Bandwidth in multi-channel systems control
channel saturation affects performance - Communication range nodes are one-hop neighbors
at one frequency but not at another - Evaluation of the overall throughput and delay
of a system with dynamic control channel
allocation