Spectrum%20Opportunity-Based%20Control%20Channel%20Assignment%20in%20Cognitive%20Radio%20Networks

1
Spectrum 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

2
The 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
3
Cooperative 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!
4
Current Practices for Control Channel Assignment

• Use unlicensed bands such as ISM bands

Already overcrowded Uncontrolled interference No
guaranteed performance
5
Current 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
6
Dynamic Control Channel Assignment

• Allocate one of the idle channels for control
• Creates the following circular dependency

7
Further 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
8
Spectrum-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
9
Neighbor 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.

10
Neighbor 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)
11
Time 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)
12
Cluster-based Control Channel Assignment

• Partition the network into clusters
• Take into account local idle channels

13
Mapping 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)
14
Mapping 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
15
Problem 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

16
Heuristic 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
17
Distributed 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)

18
Control Channel Migration
CRA
CH
CRB
PR
19
Clustering Performance
20
Further 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