On ergodic sum capacity of fading cognitive multiple-access and broadcast channels

Rui Zhang, Shuguang Cui, Ying-Chang Liang

This 2009 paper by Rui Zhang, Shuguang Cui, and Ying-Chang Liang is a foundational piece of research in the field of Cognitive Radio (CR). It explores the theoretical limits of how much data a secondary network can transmit when it "borrows" spectrum from a primary licensed network, specifically focusing on multi-user scenarios.

TL;DR

The paper derives the optimal power allocation strategies to maximize the average (ergodic) sum capacity of fading cognitive multiple-access and broadcast channels. It proves that simple scheduling—where only one user transmits at a time (D-TDMA)—is often the mathematically optimal way to maximize throughput while strictly protecting primary users from interference.

What Problem Does This Paper Address?

In Cognitive Radio, "secondary users" (SUs) are allowed to use the spectrum of "primary users" (PUs), provided they don't cause harmful interference. While previous research focused on simple point-to-point links, this paper tackles the more complex multiuser environment:

  1. Cognitive Multiple-Access Channel (C-MAC): Multiple SUs sending data to one secondary base station (BS).
  2. Cognitive Broadcast Channel (C-BC): One secondary BS sending data to multiple SUs.

The core challenge is managing two types of constraints simultaneously across a fading environment:

  • Transmit-Power Constraints (TPC): Limits on how much power the secondary transmitters can use.
  • Interference-Power Constraints (IPC): Limits on how much "noise" the secondary network can leak into the primary receivers.

The authors look at both Long-Term (LT) constraints (averaging power over time) and Short-Term (ST) constraints (strict limits at every single moment), creating a matrix of four possible scenarios for each channel type.

Methods

The authors frame the problem as a convex optimization task: maximizing the sum-rate of the secondary network. They employ several key mathematical tools:

  • Lagrange Duality: They use dual variables (often interpreted as "shadow prices" for power and interference) to decompose the complex long-term average problem into simpler sub-problems that can be solved for each individual fading state.
  • KKT Conditions: By analyzing the Karush-Kuhn-Tucker optimality conditions, they determine the structure of the optimal power allocation.
  • MAC-BC Duality: They leverage a known principle that relates the capacity of a broadcast channel to a multiple-access channel, allowing them to solve the C-BC case by transforming it into an equivalent C-MAC problem.
  • Numerical Ellipsoid Method: Used to iteratively find the optimal dual variables when closed-form solutions aren't available.

Datasets

The paper does not use real-world datasets. Instead, it relies on Monte Carlo simulations using:

  • Synthetic Fading Channels: 10,000 randomly generated channel power gain vectors.
  • Distribution: Circular Symmetric Complex Gaussian (CSCG) random variables, which model standard Rayleigh fading environments.
  • Parameters: Simulations were run for various numbers of secondary users ($K$) and primary receivers ($M$).

Key Insights and Ideas

  • The Optimality of D-TDMA: One of the most significant findings is that for the Broadcast Channel (C-BC), Dynamic Time-Division Multiple Access (D-TDMA) is optimal across all constraint types. This means the BS should simply pick the "best" user at any given moment and give them all the power.
  • The "M+1" Rule for MAC: In the Multiple-Access Channel (C-MAC) with short-term interference constraints, the authors show that the number of users who should transmit simultaneously is at most $M+1$ (where $M$ is the number of primary receivers). If there is only one primary receiver, at most two secondary users should talk at once.
  • Interference "Prices": The optimal power allocation resembles a modified "water-filling" algorithm. However, instead of just looking at the secondary user's channel quality, the "water level" is adjusted based on the "price" of interference—how much the secondary transmission hurts the primary receivers at that specific moment.
  • Multiuser Diversity vs. Interference: The paper highlights a new form of multiuser diversity. In a cognitive network, a user might be selected for transmission not just because they have a great path to the base station, but because they have a very poor path to the primary receivers, allowing them to blast more power without causing interference.

Key Results

  • Capacity Gains: Long-term constraints (LT-TPC and LT-IPC) yield significantly higher capacity than short-term constraints. This is because LT constraints allow the system to "save" its interference budget for moments when the channel to the secondary BS is excellent.
  • DRA vs. FRA: The proposed Dynamic Resource Allocation (DRA) schemes vastly outperform heuristic Fixed Resource Allocation (FRA) (like simple round-robin scheduling). In some simulated C-BC scenarios, DRA achieved nearly 4x the throughput of FRA as the number of users increased.
  • Saturation: As transmit power ($P$) increases, the capacity eventually saturates. This happens because the interference constraint (protecting the primary user) becomes the bottleneck, rendering extra transmit power useless.

Limitations and Future Work

  • Perfect CSI Assumption: The paper assumes the secondary base station has perfect Channel State Information (CSI) for all links, including the links between secondary transmitters and primary receivers. In practice, primary users rarely cooperate to provide this data.
  • SISO Focus: The study is limited to Single-Input Single-Output (SISO) nodes. The authors note that extending this to MIMO (multiple antennas) would be a natural next step.
  • Complexity: While D-TDMA simplifies things, the "optimal" calculation still requires solving dual problems that might be computationally heavy for real-time hardware with very fast fading.
  • Sensing Errors: The model assumes a "spectrum underlay" approach and does not account for the potential errors in spectrum sensing that occur in "opportunistic access" models.

Summary updated

BibTeX
@article{zhang2009ergodic,
  author = {Rui Zhang and Shuguang Cui and Ying-Chang Liang},
  title = {{On ergodic sum capacity of fading cognitive multiple-access and broadcast channels}},
  journal = {IEEE Trans. Information Theory},
  year = {2009},
  url = {https://openreview.net/forum?id=G07LxEPMRx},
}

Loading related content...

Figure