Joint Sensing and Power Allocation in Nonconvex Cognitive Radio Games: Nash Equilibria and Distributed Algorithms

Gesualdo Scutari, Jong-Shi Pang

TL;DR

This paper presents a breakthrough framework for Cognitive Radio (CR) networks by jointly optimizing how secondary users sense the spectrum and how they allocate power to transmit. By modeling this as a nonconvex game with distributed pricing mechanisms, the authors prove that decentralized users can maximize their throughput while strictly protecting primary users, outperforming traditional methods that treat sensing and transmission as separate problems.

What Problem Does This Paper Address?

In the world of wireless communications, spectrum is a scarce resource. Cognitive Radio (CR) allows "Secondary Users" (SUs) to opportunistically use frequency bands assigned to "Primary Users" (PUs) when those bands are idle. However, this creates a two-fold challenge:

  1. The Sensing-Transmission Trade-off: If an SU spends too much time sensing the spectrum to avoid interfering with a PU, it has less time left to actually send data. Conversely, sensing too quickly leads to errors and harmful interference.
  2. Decentralization vs. Global Constraints: Protecting a PU usually requires a "global" interference limit—the sum of interference from all SUs must stay below a threshold. Enforcing this typically requires a central controller, which is impractical for ad-hoc or distributed networks.

Most existing research optimizes sensing and power allocation in isolation. This paper tackles the "messy" reality where these variables are coupled, resulting in a complex, nonconvex optimization problem that classical game theory struggles to solve.

Methods

The authors model the network as a Nonconvex Nash Game. Each SU is a player trying to maximize their "opportunistic throughput" by jointly choosing three variables: sensing duration, detection thresholds, and power allocation across subcarriers.

Key technical components include:

  • Variational Inequalities (VIs): Since the game is nonconvex, the authors use the mathematical framework of VIs to analyze the existence and uniqueness of a Nash Equilibrium (NE).
  • Distributed Pricing: To enforce global interference constraints without a central manager, they introduce a pricing mechanism. SUs are "penalized" in their objective function based on their contribution to total interference. These prices are treated as additional variables to be optimized within the game.
  • Distributed Algorithms: They propose several "Best-Response" algorithms:
  • Jacobi/Simultaneous: Users update strategies at the same time.
  • Asynchronous: Users update whenever they have new info, making the system robust to delays.
  • Consensus Subroutines: SUs use limited local signaling (consensus algorithms) to estimate network-wide averages, such as the total interference or average sensing time, without needing a central hub.

Datasets

The paper is theoretical and relies on numerical simulations rather than external real-world datasets.

  • Setup: They simulated a hierarchical CR network with 2 PUs and 10–15 SUs.
  • Channel Model: Frequency-selective interference channels modeled as FIR filters (order L=10) with 1024 subchannels.
  • Benchmarks: The proposed algorithms were compared against state-of-the-art centralized Network Utility Maximization (NUM) and decentralized power-control-only schemes.

Key Insights and Ideas

  • The "Price" as a Player: One of the most elegant ideas is treating the interference price as a virtual player in the game. This allows a global constraint (protecting the PU) to be "baked into" the local decision-making process of each SU.
  • Joint is Better: The "aha moment" is the demonstration that sensing parameters and power allocation are not just related—they are mathematically inseparable for optimal performance. Optimizing them together provides a "sensing-throughput" sweet spot that isolated optimization misses.
  • Physical Interpretation of Convergence: The authors show that their mathematical conditions for a unique equilibrium have a physical meaning: they essentially require that "Multi-User Interference" (MUI) is sufficiently low compared to background noise, mirroring conditions found in simpler, convex power-control games.

Key Results

  • Throughput Gains: The joint optimization approach significantly outperformed traditional NUM-based schemes. In scenarios with stringent interference constraints, the proposed method showed a sum-throughput gain of up to 80% over methods that do not optimize sensing (see Fig. 3).
  • Fast Convergence: Despite the nonconvexity, the distributed algorithms converge in just a few iterations (typically <20), making them viable for real-time hardware implementation.
  • Robustness: The asynchronous version of the algorithm was shown to be robust against missing or outdated measurements, a critical feature for real-world wireless environments.

Limitations and Future Work

  • Low-Interference Assumption: The proofs for uniqueness and convergence rely on the system being in a "low-to-medium" interference regime. In extremely dense networks where interference is dominant, the game might have multiple equilibria or fail to converge quickly.
  • Signaling Overhead: While the consensus algorithms are "limited" signaling, they still require SUs to exchange some information. The authors acknowledge a trade-off: more signaling leads to better performance, but zero-signaling (totally uncoordinated) cases result in a performance hit.
  • Future Directions: The authors suggest extending the model to handle more complex "device-level uncertainties" (like unpredictable noise floors) and exploring "Quasi-Nash Equilibria" for cases where a standard NE cannot be guaranteed.

Summary updated

Abstract

In this paper, we propose a novel class of Nash problems for cognitive radio (CR) networks, modeled as Gaussian frequency-selective interference channels, wherein each secondary user (SU) competes against the others to maximize his own opportunistic throughput by choosing jointly the sensing duration, the detection thresholds, and the vector power allocation. The proposed general formulation allows us to accommodate several (transmit) power and (deterministic/probabilistic) interference constraints, such as constraints on the maximum individual and/or aggregate (probabilistic) interference tolerable at the primary receivers. To keep the optimization as decentralized as possible, global (coupling) interference constraints are imposed by penalizing each SU with a set of time-varying prices based upon his contribution to the total interference; the prices are thus additional variable to optimize. The resulting players' optimization problems are nonconvex; moreover, there are possibly price clearing conditions associated with the global constraints to be satisfied by the solution. All this makes the analysis of the proposed games a challenging task; none of classical results in the game theory literature can be successfully applied. The main contribution of this paper is to develop a novel optimization-based theory for studying the proposed nonconvex games; we provide a comprehensive analysis of the existence and uniqueness of a standard Nash equilibrium, devise alternative best-response based algorithms, and establish their convergence. Some of the proposed algorithms are totally distributed and asynchronous, whereas some others require limited signaling among the SUs (in the form of consensus algorithms) in favor of better performance; overall, they are thus applicable to a variety of CR scenarios, either cooperative or noncooperative, which allows the SUs to explore the existing tradeoff between signaling and performance.
BibTeX
@article{scutari2013joint,
  author = {Gesualdo Scutari and Jong-Shi Pang},
  title = {{Joint Sensing and Power Allocation in Nonconvex Cognitive Radio Games: Nash Equilibria and Distributed Algorithms}},
  journal = {IEEE Trans. Information Theory},
  year = {2013},
  url = {https://openreview.net/forum?id=gDYbrCaTOA4},
}

Loading related content...

Figure