Abstract
Cognitive Radio (CR) technology addresses the scarcity of radio spectrum by allowing secondary users to opportunistically access licensed bands when primary users are inactive. This paper introduces a spatial-spectral joint detection framework designed for wideband spectrum sensing. By leveraging the spatial diversity of multiple cognitive radios and optimizing the weighting of local sensing data across various subbands, the proposed method maximizes network throughput while adhering to strict interference constraints. The core contribution involves reformulating a complex non-convex optimization problem into a solvable convex form. Numerical results indicate that this joint optimization significantly outperforms individual sensing strategies.
Introduction
The radio spectrum is a finite resource increasingly crowded by modern wireless applications. Cognitive Radio (CR) offers a solution by identifying "spectrum holes"—frequency bands temporarily unused by licensed primary users. However, reliable detection is hampered by physical phenomena such as fading and shadowing, where a single sensor might fail to detect a primary user due to environmental obstructions. Furthermore, sensing modern wideband signals across multiple subchannels introduces significant computational complexity.
To overcome these hurdles, this research proposes a cooperative sensing framework that utilizes "spatial-spectral" diversity. By combining observations from multiple spatially distributed radios across a wide frequency range, the system can make more reliable decisions than any single device.
Methodology
The proposed system employs a cooperative sensing scheme based on the linear combination of local energy detection statistics. Each of the $N$ cognitive radios in the network monitors $K$ subchannels, transmitting its local statistics to a central fusion center.
The primary objective is to maximize the aggregate opportunistic throughput, defined as the total data rate achievable by secondary users in detected spectrum holes. This is formulated as an optimization problem (P1) subject to two main constraints:
- Interference Protection: The probability of interfering with primary users must remain below a predefined threshold.
- Spectrum Utilization: A minimum level of sensing reliability must be maintained to ensure efficient use of the available spectrum.
Because the initial formulation is non-convex and difficult to solve, the authors apply a mathematical reformulation. By utilizing a change of variables and the properties of the Q-function, they transform the problem into a convex optimization problem (P2). This transformation allows for efficient, real-time computation of optimal weights ($w$) for each sensor and subband, prioritizing radios with superior channel conditions.
Results and Discussion
The framework was evaluated using numerical simulations of a multiband OFDM system with 8 subchannels and 2 cognitive radios. The simulation parameters included specific channel gains, throughput rates, and interference costs for each subband.
As shown in Figure 2, the "Joint Optimization" approach consistently achieves higher aggregate throughput compared to individual radios (CR1 and CR2) acting independently. The results demonstrate that the system effectively maximizes data rates while strictly respecting the "interference budget" ($\epsilon$) of the primary network. Interestingly, the throughput gains eventually level off as the interference constraint is relaxed, indicating that the system becomes limited by false alarm requirements rather than interference thresholds.
Conclusion
This paper demonstrates that treating wideband spectrum sensing as a joint spatial-spectral problem significantly improves the performance of cognitive radio networks. The conversion of the sensing problem into a convex optimization task provides a practical path for implementation in real-world hardware.
Future research should address the overhead associated with the control channel used for data fusion, as well as the system's performance in highly dynamic, mobile environments. Additionally, exploring the scalability of this approach to networks with hundreds of sensors and subchannels remains an important area for investigation.
Summary updated