1. Spectrum Utilization The goals of this project: (1) Development of algorithms for the distributed detection of incumbents/trespassers in any frequency band of interest; (2) Determination of the correct response to detected incumbents (3) Testing of our algorithms on the 802.11b network in the eStadium testbed in Ross Ade Stadium at Purdue University. Abstract

2. Software and Simulation Tools Throughput measurement: iperf Linux drivers for Atheros cards –Flexibility in terms of accessing code –Cisco AP: telnet connection to set channels Several APs: –Optimizing System: UDP –Incumbent System: Generate UDP traffic on our channel –Other APs around

3. Dynamic Channel Allocation – Simple Initial Algorithm Scan channels – Received Signal Strength (RSS) EWMA filter on channel j: RSS j =  RSS j k-1 +(1-  ) RSS j k Weight RSS on channel i from neighboring channels j Find the channel with minimum interference power Communicate with AP to switch to that channel

4. Centralized versus Distributed Spectrum Sensing Centralized ProcessingDistributed Processing

5. Taxonomy of Spectrum Sensing Techniques for CR Transmitter Detection Matched Filter Detection Energy Detection Cyclostationary Feature Detection Interference Based Detection Cooperative Detection

6. Initial Theoretical Work on Distributed Spectrum Sensing – Matched Filter Consider a system of N radios, each employing matched filter technique for spectrum sensing. Hypothesis at each radio: X k is the received signal, s k is signal that is to be detected, h is the channel gain for the received signal and n k is N (0,  2 ) Receivers use Neyman-Pearson Criteria for the hypothesis testing problem

7. Performance of Detection Systems – Matched Filter Non-collaborative and collaborative scheme under the AWGN channel model Collaborative scheme: –Final decision is based on k out of N rule: k = 1 refers to OR decision fusion rule k = N corresponds to AND fusion rule k = N/2 corresponds to MAJORITY fusion rule

8. Experimental results from measurement tests System Setup -Two distances 50 and 250 feet -At each distance RTS/CTS was on for both and off for both systems -At each RTS/CTS setting the packets lengths were 10, 20 and 1470 bytes -For each packet length iperf was ran. The incumbent system was ran before and after the switching channel system ran. Experimental Description

9. Effect of Packet Length and RTS/CTS on Throughput at a distance of 50 feet

10. Effect of Distance on Throughput with packet length of 1470 Bytes