Transmit without Regrets: Online Optimization in MIMO-OFDM Cognitive Radio Systems

Panayotis Mertikopoulos, Elena Veronica Belmega

TL;DR

This paper introduces Augmented Exponential Learning (AXL), a distributed algorithm that allows secondary wireless users to maximize their data rates in dynamic "cognitive radio" environments. By utilizing "no-regret" optimization, the algorithm ensures that users perform as well as the best possible fixed strategy in hindsight, even when faced with rapidly changing conditions and significant measurement errors.

What Problem Does This Paper Address?

The radio spectrum is a finite and crowded resource. Cognitive Radio (CR) systems attempt to solve this by creating a hierarchy: "Primary Users" (PUs) own the rights to a frequency, while "Secondary Users" (SUs) "free-ride" on unused gaps (white spaces).

The challenge is that this environment is incredibly volatile. PUs come and go unpredictably, and signal quality fluctuates due to physical movement (fading). Traditional game-theory concepts, like the Nash Equilibrium, assume a static environment where everyone settles into a stable state. In the real world, by the time an SU calculates an "equilibrium," the environment has already changed. SUs need a way to optimize their transmission profiles—how much power to send over which frequency and in which spatial direction—"on the fly" using only local, often imperfect, information.

Methods

The authors propose a framework called Online Optimization, specifically focusing on Regret Minimization. Instead of trying to predict the future, a "no-regret" policy ensures that, over time, the user's average performance is at least as good as the best fixed strategy they could have chosen if they had known the entire history of the network.

The core contribution is the Augmented Exponential Learning (AXL) algorithm. It decomposes the complex task of managing a MIMO-OFDM system (Multiple-Input Multiple-Output combined with Orthogonal Frequency Division Multiplexing) into two parts:

  1. Power Allocation (OFDM): Deciding how to distribute power across different frequency subcarriers.
  2. Signal Covariance (MIMO): Deciding the spatial direction of the signal across multiple antennas.

AXL uses Matrix Exponential Learning, a technique rooted in "online mirror descent." It assigns "scores" to different transmission strategies based on past performance and updates them exponentially. Crucially, the algorithm is:

  • Distributed: Each user only needs their own local data.
  • Stateless: Users don't need to know the network topology or what others are doing.
  • Robust: It is mathematically proven to work even if the Channel State Information (CSI) measurements are noisy or imperfect.

Datasets

The paper relies on extensive numerical simulations rather than static external datasets, as is standard for wireless communications research.

  • Channel Model: They used the Jakes model to simulate Rayleigh fading, which realistically mimics signal fluctuations for users moving at pedestrian speeds (3–5 km/h).
  • Network Setup: A simulated environment with 10 Primary Users and 40 Secondary Users communicating over 256 subcarriers.
  • Traffic Patterns: PU arrivals and departures were modeled using a Poisson process, creating a stochastic, "bursty" environment.

Key Insights and Ideas

  • The "No-Regret" Philosophy: In a world that changes too fast for equilibrium, "not having regrets" is a more robust goal than "finding the optimum."
  • Robustness to Imperfect Data: One of the most surprising findings is that AXL maintains its "no-regret" properties even with "arbitrarily large" observation errors, provided there is no systematic bias. This makes it far more practical than traditional "water-filling" algorithms, which often fail when data is noisy.
  • Tracking the Moving Target: The simulations show that AXL doesn't just settle for a safe average; it actually "tracks" the theoretical maximum achievable rate remarkably closely as the environment evolves.

Key Results

  • Convergence Rate: The authors proved that AXL achieves a worst-case average regret of $O(T^{-1/2})$ after $T$ time steps. This is a "tight" bound, meaning it's the best performance one can theoretically guarantee in such an environment.
  • Superiority over Water-Filling: In head-to-head simulations, AXL reached maximum system capacity within 3–4 iterations. In contrast, standard "Simultaneous Water-Filling" (SWF) failed to converge entirely in many dynamic scenarios.
  • Performance under Noise: When 50% relative error was introduced into the channel measurements, AXL still achieved system capacity (albeit more slowly), while traditional methods offered no improvement over a random initial guess.

Limitations and Future Work

  • Extreme Mobility: While AXL handles pedestrian speeds well, the authors note that at very high velocities (e.g., high-speed trains), the channel might change faster than the algorithm can update, potentially degrading performance.
  • Adaptive Regret: The paper focuses on "external regret" (comparison to a fixed strategy). The authors suggest that "adaptive regret"—which compares performance against a strategy that can change over time—might be a better metric for even more volatile environments.
  • Energy Awareness: The current model assumes users want to maximize data rate at full power. Future iterations could incorporate "energy-aware" objectives where users balance speed against battery life.

Summary updated

Abstract

In this paper, we examine cognitive radio systems that evolve dynamically over time due to changing user and environmental conditions. To combine the advantages of orthogonal frequency division multiplexing (OFDM) and multiple-input, multiple-output (MIMO) technologies, we consider a MIMO-OFDM cognitive radio network where wireless users with multiple antennas communicate over several non-interfering frequency bands. As the network's primary users (PUs) come and go in the system, the communication environment changes constantly (and, in many cases, randomly). Accordingly, the network's unlicensed, secondary users (SUs) must adapt their transmit profiles "on the fly" in order to maximize their data rate in a rapidly evolving environment over which they have no control. In this dynamic setting, static solution concepts (such as Nash equilibrium) are no longer relevant, so we focus on dynamic transmit policies that lead to no regret: specifically, we consider policies that perform at least as well as (and typically outperform) even the best fixed transmit profile in hindsight. Drawing on the method of matrix exponential learning and online mirror descent techniques, we derive a no-regret transmit policy for the system's SUs which relies only on local channel state information (CSI). Using this method, the system's SUs are able to track their individually evolving optimum transmit profiles remarkably well, even under rapidly (and randomly) changing conditions. Importantly, the proposed augmented exponential learning (AXL) policy leads to no regret even if the SUs' channel measurements are subject to arbitrarily large observation errors (the imperfect CSI case), thus ensuring the method's robustness in the presence of uncertainties.
BibTeX
@article{mertikopoulos2014transmit,
  author = {Panayotis Mertikopoulos and Elena Veronica Belmega},
  title = {{Transmit without Regrets: Online Optimization in MIMO-OFDM Cognitive Radio Systems}},
  journal = {IEEE J. Sel. Areas Commun.},
  year = {2014},
  url = {https://openreview.net/forum?id=ol5A7xctdiT},
}

Loading related content...

Figure