Twitter Feed
Couldn't connect with Twitter

What are you looking for?

Simply enter your keyword and we will help you find what you need.

Spectrum Sensing in Cognitive Radio Networks using

Background
Spectrum sensing is becoming an increasingly important function in cognitive radio networks. Hence, spectrum sensing enables secondary users to efficiently identify and utilize spectrum holes without causing interference to primary users. Different forms of spectrum sensing are there but these forms have their own limitations.

Challenges
One such form of spectrum sensing is non-cooperative sensing which does not provide accurate enough results by the use of deep learning algorithms. Therefore to provide accurate results, there is a need for a spectrum sensing form that increases reliability in determining the presence of PUs due to the advancements in DRL methods that have shown promising results in the field of communication networks.

Goal
To deal with the limitations of deep learning algorithms for both spectrum sensing scenarios, this project aims to propose an offline DRL algorithm that significantly improves the detection accuracy of PU.

Proposed Methodology
To improve the detection accuracy of PU in both scenarios, an offline DRL-based spectrum sensing algorithm named the conservative Q learning method for CRN is proposed. For this purpose, a discrete version of CQL is utilized where firstly the task of local spectrum sensing is formulated. To evaluate the presence of PU in this phase, each SU employs DCQL models for the various signal-to-noise ratio (SNR) regions. Then, utilizing the present and historical energy levels, each SU provides a one-bit result. These findings are subsequently forwarded to the FC, where the CSS task is established. Another DCQL agent is used at the FC, which accepts all one-bit findings as input and outputs the final judgment about the presence or absence of the PU. The SUs are then informed of the final decision.

Furthermore, to mimic the real-world wireless application scenarios, discrete and continuous-time models for PU spectrum utilization, along with different PU activity patterns are taken under consideration. In the end, simulation is performed to compare the proposed method with other benchmarks for both LSS and CSS scenarios.

Results
Different traffic loads for DTMC and CTMC are simulated in both settings to demonstrate the usefulness of the proposed strategy. In the context of LSS, the suggested method yields performance comparable to state-of-the-art LSTM while requiring 2x less model computational time. In the scenario of cooperative spectrum sensing, the suggested method outperforms the stateof-the-art DCS-SD while decreasing transmission overheads and model computation time significantly. The figure below shows the computation time of PU detection methods in the case of local spectrum sensing.

Conclusion
From the findings, it is evident that the proposed method outperformed the state-of-the-art deep cooperative spectrum method in terms of sensing errors.

 

 

Previous Project