Optimal hard fusion strategies for cognitive radio networks

Abstract

Abstract-Optimization of hard fusion spectrum sensing using the k-out-of-N rule is considered. Two different setups are used to derive the optimal k. A throughput optimization setup is defined by minimizing the probability of false alarm subject to a probability of detection constraint representing the interference of a cognitive radio with the primary user, and an interference management setup is considered by maximizing the probability of detection subject to a false alarm rate constraint. It is shown that the underlying problems can be simplified to equality constrained optimization problems and an algorithm to solve them is presented. We show the throughput optimization and interference management setups are dual. The simulation results show the majority rule is optimal or near optimal for the desirable range of false alarm and detection rates for a cognitive radio network. Furthermore, an energy efficient setup is considered where the number of cognitive radios is to be minimized for the AND and the OR rule and a certain probability of detection and false alarm constraint. The simulation results show that the OR rule outperforms the AND rule in terms of energy efficiency.

Low-Complexity Optimal Hard Decision Fusion Under the Neyman–Pearson Criterion

“…In general, the constrained optimization of the NP criterion increases the problem complexity. For problems with monotonic property, low complexity methods like bisection, gradient descent etc., can be used to compute the optimal threshold in some cases [19][20][21]. The non-monotonic property of the LR complicates the optimal decision equation, which is generally intractable in C-OS [12].…”

“…As a result the maximum value of | 90 and weight 7 kg is obtained when items {0, 2} are placed in the box. 19) with the threshold λ chosen as 3.0 < λ < 10.0 for this example.…”

“…The rules {MSB-2, MCB-2 } of {semi-blind, completely-blind} categories use Bayesian inference to estimate the unknown value and outperform for most of the cases in their respective categories. Further avenues of research include the derivation of blind rules for more advanced cooperative/collaborative spectrum sensing schemes [52].It is shown in chapter 2 that the optimal solution for the GDFP under the Neyman-Pearson criterion can be computed using low complexity methods like bisection, gradient descent etc., in some cases[19][20][21] if the LR function is monotonic. However for the non-monotonic problems, the optimal fusion rule requires multi-threshold decision equation and the computations require exponentially complex exhaustive searchmethods [1, 2, 51].Secondly, it is shown that the non-randomized hard decision fusion problem is in the form of the classical 0 − 1 Knapsack problem (KP) and thereby a low complexity solution using dynamic programming (DP) is proposed.…”

EESS: An Energy‐Efficient Spectrum Sensing Method by Optimizing Spectrum Sensing Node in Cognitive Radio Sensor Networks

“…To this end, there are some existing node selection methods in various works [3,[22][23][24][25][26][27]. Specifically, an optimal hard fusion strategy was proposed to maximize the energy efficiency in [22]. In [23], an optimal number of multihop-based SUs was derived.…”

Optimal Number of Cognitive Users in $K$ -Out-of- $M$ Rule

“…In [16], optimal K is found to maximize the CUs network throughput while satisfying protection constraint to the PU. In [17], an algorithm for the optimal K is presented to maximize the global detection probability subject to a constraint on global false alarm probability. In [18], the optimal K and the optimal detection threshold of the multi-hop CR network are derived.…”