Cluster Size Optimization in Cooperative Spectrum Sensing

In this paper, we study and optimize the cooperation cluster size in cooperative spectrum sensing to maximize the throughput of secondary users (SUs). To calculate the effective throughput, we assume each SU spends just 1 symbol to negotiate with the other SUs in its transmission range. This is the minimum overhead required for each SU to broadcast its sensing decision to the other members of the cluster. When the number of SUs is large, the throughput spent for the negotiation is noticeable and therefore increasing the cooperation cluster size does not improve the effective throughput anymore. In this paper, we calculate the effective throughput as a function of the cooperation cluster size, and then we maximize the throughput by finding the optimal cluster size. Various numerical results show that when decisions are combined by the OR-rule, the optimum cooperation cluster size is less than when the AND-rule is used. On the other hand, the optimum cluster size monotonically decreases with the increase in the average SNR of the SUs. Another interesting result is that when the cluster size is optimized the OR-rule always outperforms the AND-rule.

💡 Research Summary

This paper investigates the trade‑off between cooperative sensing gain and control‑overhead in cognitive radio networks, focusing on how the size of a cooperation cluster influences the effective throughput of secondary users (SUs). The authors adopt a minimal‑overhead model in which each SU spends exactly one symbol (a fixed time slot) to broadcast its binary sensing decision to all other SUs within its transmission range. Consequently, a cluster containing N SUs incurs an overhead of N·τ symbols, where τ denotes the fraction of a frame occupied by a single symbol. The data transmission phase therefore occupies only (1 – N·τ) of the frame.

Individual SUs perform energy detection, characterized by a detection probability Pd and a false‑alarm probability Pfa that depend on the average signal‑to‑noise ratio (SNR). The paper considers two classic decision‑fusion rules: the OR‑rule (the channel is declared occupied if any SU reports “busy”) and the AND‑rule (the channel is declared occupied only if all SUs report “busy”). Under these rules the combined false‑alarm probabilities become Pfa,OR = 1 – (1 – Pfa)^N and Pfa,AND = (Pfa)^N, respectively. The effective throughput η(N) is modeled as

\