1. COGNITIVE RADIO FOR NEXT-GENERATION WIRELESS NETWORKS: AN APPROACH TO OPPORTUNISTIC CHANNEL SELECTION IN IEEE BASED WIRELESS MESH
Dusit Niyato, Nanyang Technological University
Ekram Hossain, University of Manitoba
IEEE Wireless Communication Feb. 2009

2. Outline Introduction Cognitive Radio
Basic Components, Approaches
In Different Wireless Systems
Research Issues in Protocol Design
An Approach to Opportunistic Channel Selection in IEEE Based Wireless Mesh
System Model
Dynamic Opportunistic Channel Selection Scheme
Performance Evaluation
Conclusion
Comments

3. Introduction
Frequency spectrum is the scarcest resource for wireless communications
may become congested to accommodate diverse types of air interfaces in next-generation wireless networks
Software radio
Improves the capability of a wireless transceiver by using embedded software
Enable the radio transceiver to operate in multiple frequency bands
Cognitive radio
A special type of software defined radio
Able to intelligently adapt itself to the changing environment

4. Cognitive Radio Basic Components Observation Process Learning Process
Measurement and noise reduction mechanism
Passive observation
The radio transceiver silently listens to the environment.
Active observation
Special messages or signals are transmitted and measured to obtain information about the surrounding environment
Learning Process
Extract useful information from collected data
Reinforcement learning algorithm
is used when the correct solution is unknown
Learning through interactions

5. Cognitive Radio (cont’d)
Planning and Decision Making Process
Using knowledge obtained from learning to schedule and prepare for the next transmission
A transceiver must decide to choose the best strategy to achieve the target objective
Action
The action of a transceiver is controlled by the planning and decision making process

6. Cognitive Radio (cont’d)
Approaches
Estimation Technique
Obtain information about the ambient network environment
Game Theory
Evolutionary Computation
Genetic algorithm
Fuzzy Logic
Markov Decision Process
Pricing Theory
Theory of Social Science
Reinforcement Learning

7. Cognitive Radio (cont’d)
In different wireless systems
IEEE and Networks
May operate in the same unlicensed frequency band
Efficient spectrum management and planning are required
IEEE Networks (WRANs)
The first wireless communication standard adopting intelligent software defined radio
Ultra Wideband-based (WPANs)
Cooperative Diversity Wireless Networks
Primary users and secondary users

8. Cognitive Radio (cont’d)
Research issues in protocol design
Lightweight and cooperative protocols for cognitive radio networks
Battery-limited, energy consumption for the execution of estimation, learning, and decision making algorithm should be minimized
Cross-layer optimization in cognitive radio networks
To optimize QoS performance in a cognitive radio network

9. Dynamic Channel Selection Scheme
In the proposed scheme
A wireless node/mesh client learns physical (i.e., signal strength) and MAC layer (i.e., collision probability)
Accordingly selects the best channel to connect to a mesh router
The decision can be made independently in each node in a distributed manner by using an intelligent algorithm

10. Dynamic Channel Selection Scheme (cont’d)
System Model
IEEE Mesh
100m*100m
No centralized controller

11. Dynamic Channel Selection Scheme (cont’d)
Fuzzy logic controller
Pc(f): collision probability on channel f  estimate the amount of traffic load
γf: estimated signal strength

12. Dynamic Channel Selection Scheme (cont’d)
Wireless node utility
The decision on dynamic channel selection at each node is based on utility function of collision probability Pc(f) and received signal strength γf on channel f.
Both collision probability and received signal strength impact the throughput and error performances experienced by a wireless node.

13. Dynamic Channel Selection Scheme (cont’d)
Fuzzy logic
Use “collision probability” as an indicator of traffic load in each channel
The interference rules are used to gain information on the traffic load condition in a channel
Estimated collision prob.
Result utility
Example:

14. Dynamic Channel Selection Scheme (cont’d)
Let mf,i denote the membership function for channel f obtained from fuzzification.
This mf,i can be obtained using a standard fuzzification method.
Then the fitness of rule k to the traffic load condition can be obtained from
The estimated utility
The normalized fitness

15. Dynamic Channel Selection Scheme (cont’d)
Learning algorithm
is used to approximate the utility Ui,f,k perceived by each wireless node corresponding to the different traffic load condition in the service area
α: the learning rate
Uoldi,f,k: the utility of the previous learning iteration

16. Dynamic Channel Selection Scheme (cont’d)
Decision on Channel Selection
Wireless node i chooses channel that provides the highest Ui,f
This channel selection is executed periodically.
The decision can be made if the estimated collision probability and received signal strength change by an amount larger than the predefined thresholds,
which implies that one or more new nodes are accessing the channel and/or some nodes have terminated connections with the corresponding mesh router.

17. Dynamic Channel Selection Scheme (cont’d)
Performance Evaluation
Each router operates in DCF mode
For the channel selection scheme we set α: 0.1, and it is executed at each node periodically every 2 min.
Using MATLAB to run the time-driven simulation

18. Dynamic Channel Selection Scheme (cont’d)
Wireless nodes and the associated mesh routers: a) at time 0; b) after 30 minutes

19. Dynamic Channel Selection Scheme (cont’d)
Wireless nodes and the associated mesh routers (for non-uniform node distribution)
Variation in average node throughput

20. Dynamic Channel Selection Scheme (cont’d)
Effect of uniformity of node distribution on the network utility

21. Conclusion
An overview of the difference components in cognitive radio and the related approaches have been presented
The dynamic channel selection for opportunistic spectrum access in IEEE based multichannel wireless mesh networks
It performs significantly better than some of the traditional schemes, especially with non-uniform node distribution in the service area.

22. Comments Provide an introduction to cognitive radio’s approaches.
Learning rate selection is a issue.
Convergence and performance
Comparison with other channel selection scheme?