كشف الاهداف خارج الاحداثيات الشبكية في شبكات الاستشعار اللاسلكية باستخدام التعلم البايزي المتناثر (SBL) == Off-Grid Target Detection in Wireless Sensor Networks via SBL
Author name:
مصطفى عبد الرحمن جبار
Supervisor name:
احمد محمد زكي الحلى
General topic:
Electrical, Electronic and Communications Engineering
Specific topic:
Communications Engineering
Degree:
Master
University:
Al-furat Al-Awsat Technical University - Technical Engineering College - Najaf
Language:
English
University location:
Najaf
Key words:
- Energy-efficient
- Off-Grid Target Detection
- Target detection
- Compressive sensing
- Wireless Sensor Networks
- Sparse Bayesian Learning
- Adaptive Sparse Bayesian Learning
First pages:
T108280 - p.pdf
Abstract:
Energy consumption is a significant challenge in Wireless Sensor Networks (WSNs). Compressed Sensing (CS) effectively reduces energy consumption but faces limitations, such as detecting a limited number of targets and assuming targets are on-grid. This thesis proposed using Sparse Bayesian Learning (SBL) for target detection in WSNs. SBL provided a model for off-grid target detection and also enhanced target detection accuracy and achieved nearly double the target detection compared to other methods like Basis Pursuit (BP). The SBL method successfully detected up to 30 targets with very high detection performance. In comparison, the BP method detected only 15 targets under the same environmental conditions. Two variables are considered in off-grid scenarios: displacement along the x-axis and y-axis. The Off-grid detection results demonstrate that the proposed approach, which incorporates dependent assumptions, can effectively localize targets within the framework of theoretical assumptions. This is in contrast to an approach based on independent assumptions, which may not achieve the same level of accuracy. Further, this thesis proposed an adaptive Sparse Bayesian Learning (A-SBL) approach to increasing the sensors’ lifetime in wireless sensor networks (WSNs). By combining the Bayesian model with adaptive compressive sensing (A-CS), the methodology minimizes the number of sensors required for successful target detection. Initially, a few sensors are selected randomly, and the Cluster Head (CH) then calls sensors that provide maximal information, achieving the greatest error reduction. This approach enhances resource use and energy efficiency, improving overall network performance. Results confirm that this method significantly reduces energy consumption v compared to other approaches, especially with fewer targets, contributing to advancements in WSN technology. Where The results show that the energy consumption of the A-SBL approach was 53% of the energy consumption by the traditional method when detecting 5 targets. As the number of targets increased to 15 and 30, the energy consumption dropped to 30% and 17%, respectively, compared to the energy consumed by the traditional method. Additionally, when the initial sensors were increased to 15 instead of 5, the proposed method’s energy consumption was 48%, 28%, and 17% for 5, 15, and 30 targets, respectively
Full text:
0a452e04e6.pdf
