Abstract
The rapid growth of services offered through wireless communication has as consequence an increase on the demand for electromagnetic radio frequency spectrum, which is a scarce resource, mainly assigned to license holders on a long-term basis for large geographical regions, causing a large portion of the spectrum to remain unused for a significant percentage of the time. A new paradigm -- to overcome this apparent spectrum shortage -- that consists of radio devices with the ability to adapt to their spectral environment and are therefore able to make use of the available spectrum in an opportunistic manner was put forward, i.e. the Cognitive Radio paradigm.
Spectrum sensing is the key mechanism in enabling spectrum awareness in Cognitive Radio. The performance of the spectrum sensing depends on the local channel conditions, such as the multipath, shadowing and the receiver uncertainty issues. The conjunction of these conditions can result in regimes where the signal strength is below the detection threshold of the sensor, resulting in missed detections.
To overcome this limitation, there have been several proposals made in the research community towards the use of cooperation in spectrum sensing.
Since the signal strength varies with the sensor location, the worst fading conditions can be avoided if multiple sensors in different spatial locations share their local sensing measurements, i.e. take advantage of the spatial diversity.
In this thesis a Cooperative Spectrum Sensing mechanism is proposed. While identifying the key components needed to enable such mechanism, an analysis is presented regarding the correctness of the class of protocols which enable the Cooperative Spectrum Sensing mechanism. This is done by proposing and employing a process calculus variant of the $\pi$-calculus, denoted as Bounded Broadcast Calculus. This analysis is done over centralized, decentralized and relay aided topologies. The outcome of this analysis is a theorem where it is stated, which properties a protocol should have so that it can be deemed correct, i.e. that it performs as intended, over each of the considered network topologies.
The performance of data fusion schemes based on counting rules is analyzed, which lead to the proposal of an adaptive counting rule algorithm that adapts the decision threshold according to the performance of the local detectors. A study is done about the impact of using local detectors in the data fusion scheme which are experiencing different channel conditions, i.e. some of these local detectors are experiencing the channel as free while others as occupied. In this analysis it is measured what is the impact of using data fusion in these cases, and whether it improves the detection of the resources available. Based on these insights a cluster based data fusion algorithm is proposed, which uses the correlation measured between the decisions of the local detectors over time to group the local detectors together in different clusters, and then apply the adaptive counting rule data fusion algorithm separately to each of the defined clusters. It was observed, in the case of a single signal source, that the proposed algorithm performance in regards to identified spectrum resources is almost the same as the theoretical maximum, and superior to the case where the local detectors are not divided in clusters.
Finally, a node selection mechanism which assigns the local detectors to the channel that most likely will be experienced vacant by the local detector is proposed. The purpose of using such scheme is twofold, one is to ensure that the correct amount of the local detectors is sensing each channel; the other is to increase the probability of the network finding a channel available to be used. This is accomplished by assigning the local detectors to channels that have a higher probability of being available and where most likely the local detector is outside the range of the signal source. The node selection scheme is proposed in a centralized and in a decentralized version. These versions can complement each other and therefore lead to a more robust cooperative spectrum sensing mechanism.
Spectrum sensing is the key mechanism in enabling spectrum awareness in Cognitive Radio. The performance of the spectrum sensing depends on the local channel conditions, such as the multipath, shadowing and the receiver uncertainty issues. The conjunction of these conditions can result in regimes where the signal strength is below the detection threshold of the sensor, resulting in missed detections.
To overcome this limitation, there have been several proposals made in the research community towards the use of cooperation in spectrum sensing.
Since the signal strength varies with the sensor location, the worst fading conditions can be avoided if multiple sensors in different spatial locations share their local sensing measurements, i.e. take advantage of the spatial diversity.
In this thesis a Cooperative Spectrum Sensing mechanism is proposed. While identifying the key components needed to enable such mechanism, an analysis is presented regarding the correctness of the class of protocols which enable the Cooperative Spectrum Sensing mechanism. This is done by proposing and employing a process calculus variant of the $\pi$-calculus, denoted as Bounded Broadcast Calculus. This analysis is done over centralized, decentralized and relay aided topologies. The outcome of this analysis is a theorem where it is stated, which properties a protocol should have so that it can be deemed correct, i.e. that it performs as intended, over each of the considered network topologies.
The performance of data fusion schemes based on counting rules is analyzed, which lead to the proposal of an adaptive counting rule algorithm that adapts the decision threshold according to the performance of the local detectors. A study is done about the impact of using local detectors in the data fusion scheme which are experiencing different channel conditions, i.e. some of these local detectors are experiencing the channel as free while others as occupied. In this analysis it is measured what is the impact of using data fusion in these cases, and whether it improves the detection of the resources available. Based on these insights a cluster based data fusion algorithm is proposed, which uses the correlation measured between the decisions of the local detectors over time to group the local detectors together in different clusters, and then apply the adaptive counting rule data fusion algorithm separately to each of the defined clusters. It was observed, in the case of a single signal source, that the proposed algorithm performance in regards to identified spectrum resources is almost the same as the theoretical maximum, and superior to the case where the local detectors are not divided in clusters.
Finally, a node selection mechanism which assigns the local detectors to the channel that most likely will be experienced vacant by the local detector is proposed. The purpose of using such scheme is twofold, one is to ensure that the correct amount of the local detectors is sensing each channel; the other is to increase the probability of the network finding a channel available to be used. This is accomplished by assigning the local detectors to channels that have a higher probability of being available and where most likely the local detector is outside the range of the signal source. The node selection scheme is proposed in a centralized and in a decentralized version. These versions can complement each other and therefore lead to a more robust cooperative spectrum sensing mechanism.
| Original language | English |
|---|---|
| Print ISBNs | 978-87-92328-79-3 |
| Publication status | Published - 31 Aug 2012 |
Keywords
- Spectrum Sensing
- Cooperative Spectrum Sensing
- Process calculus
- Bounded Broadcast Calculus
- Adaptive Counting Rule
- Node Selection Mechanism
- Correlation
- Data Fusion
- Clustering
- Network Topology
- Capacity Assessment
- Cognitive Radio Technology