A Cooperative Bayesian Nonparametric Framework for Primary User Activity Monitoring in Cognitive Radio Network

Walid Saad, Zhu Han, H. Vincent Poor, Tamer Basar, Ju Bin Song

Abstract

This paper introduces a novel approach that enables a number of cognitive radio devices that are observing the availability pattern of a number of primary users (PUs), to cooperate and use Bayesian nonparametric techniques to estimate the distributions of the PUs' activity pattern, assumed to be completely unknown. In the proposed model, each cognitive node may have its own individual view on each PU's distribution, and, hence, seeks to find partners having a correlated perception. To address this problem, a coalitional game is formulated between the cognitive devices and an algorithm for cooperative coalition formation is proposed. It is shown that the proposed coalition formation algorithm allows the cognitive nodes that are experiencing a similar behavior from some PUs to self-organize into disjoint, independent coalitions. Inside each coalition, the cooperative cognitive nodes use a combination of Bayesian nonparametric models such as the Dirichlet process and statistical goodness of fit techniques in order to improve the accuracy of the estimated PUs' activity distributions. Simulation results show that the proposed algorithm significantly improves the estimates of the PUs' distributions and yields a performance advantage, in terms of reduction of the average achieved Kullback-Leibler distance between the real and the estimated distributions, reaching up to 36.5% relative the non-cooperative estimates. The results also show that the proposed algorithm enables the cognitive nodes to adapt their cooperative decisions when the actual PUs' distributions change due to, for example, PU mobility.

Introduction

In Cognitive Radio Networks (CRNs), unlicensed Secondary Users (SUs) opportunistically transmit data on spectrum bands reserved for licensed Primary Users (PUs). To avoid causing harmful interference, SUs must accurately detect when a PU is active. This requires knowledge of the underlying statistical distribution of PU activity, which is often sporadic, localized, and completely unknown a priori.

Individual cognitive nodes often struggle to estimate these distributions because they only possess localized observations, which can be skewed by distance, obstacles, and signal fading. While cooperative sensing is a known solution, it typically assumes the activity distribution is already known. This research addresses the challenge of cooperatively learning and estimating unknown PU distributions from scratch while managing communication overhead.

Proposed Cooperative Framework

The authors propose a distributed algorithm for Cognitive Pilot Channel (CPC) nodes—devices dedicated to monitoring the spectrum. The framework is built on three pillars: coalitional game theory, statistical validation, and Bayesian nonparametric inference.

Illustration of the proposed cooperative framework where CPC nodes form coalitions to estimate PU activity.

1. Coalitional Game Theory

Nodes participate in a "hedonic coalition formation game." Instead of sharing data globally, which is costly and potentially inaccurate due to differing local views, nodes autonomously form disjoint, independent coalitions. They join or leave groups based on a utility function that balances the reward of improved estimation accuracy against the cost of communication overhead. This ensures the network reaches a "Nash-stable" state where no node has an incentive to unilaterally change its coalition.

2. Goodness-of-Fit Testing

To prevent merging incompatible data, nodes utilize a two-sample Kolmogorov-Smirnov (KS) test. This statistical check ensures that a potential partner's data originates from the same underlying distribution, confirming that both nodes are experiencing similar PU behavior before they cooperate.

3. Bayesian Nonparametrics (Dirichlet Processes)

Once a coalition is formed, nodes combine their data using a Dirichlet process. This Bayesian approach is "model-free," allowing the system to infer the unknown distribution's shape directly from the data without forcing it into a predefined mathematical model (like a Gaussian distribution). This flexibility is vital for adapting to diverse PU behaviors, such as those of mobile phones versus broadcast towers.

Results and Performance

The framework was evaluated using simulated CRN environments on a 2km x 2km grid. Key findings include:

  • Significant Accuracy Gains: The cooperative approach reduced the Kullback-Leibler (KL) distance—a measure of the difference between estimated and actual distributions—by up to 36.5% compared to non-cooperative estimation.
  • Rapid Convergence: The self-organizing algorithm is computationally efficient, typically reaching a stable coalition structure in approximately 7 iterations for a 30-node network.
  • Dynamic Adaptability: The system effectively handles PU mobility. When PUs move (simulated at speeds up to 100 km/h), nodes dynamically reform coalitions to maintain accurate estimates, demonstrating high resilience to environmental changes.

Conclusion

This paper presents a robust, self-organizing framework that allows cognitive nodes to cooperatively learn unknown primary user activity patterns. By combining game theory with Bayesian nonparametrics, the system achieves high estimation accuracy with low communication overhead. Future research directions include extending the model to estimate the duration of PU activity and applying these techniques to other distributed sensing problems like wireless weather monitoring.

Summary updated

Abstract

This paper introduces a novel approach that enables a number of cognitive radio devices that are observing the availability pattern of a number of primary users(PUs), to cooperate and use \emph{Bayesian nonparametric} techniques to estimate the distributions of the PUs' activity pattern, assumed to be completely unknown. In the proposed model, each cognitive node may have its own individual view on each PU's distribution, and, hence, seeks to find partners having a correlated perception. To address this problem, a coalitional game is formulated between the cognitive devices and an algorithm for cooperative coalition formation is proposed. It is shown that the proposed coalition formation algorithm allows the cognitive nodes that are experiencing a similar behavior from some PUs to self-organize into disjoint, independent coalitions. Inside each coalition, the cooperative cognitive nodes use a combination of Bayesian nonparametric models such as the Dirichlet process and statistical goodness of fit techniques in order to improve the accuracy of the estimated PUs' activity distributions. Simulation results show that the proposed algorithm significantly improves the estimates of the PUs' distributions and yields a performance advantage, in terms of reduction of the average achieved Kullback-Leibler distance between the real and the estimated distributions, reaching up to 36.5% relative the non-cooperative estimates. The results also show that the proposed algorithm enables the cognitive nodes to adapt their cooperative decisions when the actual PUs' distributions change due to, for example, PU mobility.
BibTeX
@article{saad2012cooperative,
  author = {Walid Saad and Zhu Han and H. Vincent Poor and Tamer Basar and Ju Bin Song},
  title = {{A Cooperative Bayesian Nonparametric Framework for Primary User Activity Monitoring in Cognitive Radio Network}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2012},
  url = {https://openreview.net/forum?id=F9au9_yHLB},
}
Figure