[Paper Review] On Achievable Network Capacity and Throughput-Achieving Policies over Markov ON/OFF Channels

Chih‐Ping Li, Michael J. Neely|arXiv (Cornell University)|Mar 13, 2010

Age of Information Optimization10 citations

TL;DR

This paper establishes inner and outer bounds on the capacity region of a multi-user wireless downlink with Markov ON/OFF channels where instantaneous channel states are unknown. It proposes a queue-dependent scheduling policy that stabilizes the network for all rates within the inner bound, using frame-based Lyapunov drift and stochastic coupling to prove tightness in symmetric, large-user regimes.

ABSTRACT

We study the fundamental network capacity of a multi-user wireless downlink under two assumptions: (1) Channels are not explicitly measured and thus instantaneous states are unknown, (2) Channels are modeled as ON/OFF Markov chains. This is an important network model to explore because channel probing may be costly or infeasible in some contexts. In this case, we can use channel memory with ACK/NACK feedback from previous transmissions to improve network throughput. Computing in closed form the capacity region of this network is difficult because it involves solving a high dimension partially observed Markov decision problem. Instead, in this paper we construct an inner and outer bound on the capacity region, showing that the bound is tight when the number of users is large and the traffic is symmetric. For the case of heterogeneous traffic and any number of users, we propose a simple queue-dependent policy that can stabilize the network with any data rates strictly within the inner capacity bound. The stability analysis uses a novel frame-based Lyapunov drift argument. The outer-bound analysis uses stochastic coupling and state aggregation to bound the performance of a restless bandit problem using a related multi-armed bandit system. Our results are useful in cognitive radio networks, opportunistic scheduling with delayed/uncertain channel state information, and restless bandit problems.

Motivation & Objective

• To characterize the fundamental network capacity of a multi-user downlink when channel states are not directly measurable and are modeled as Markov ON/OFF processes.
• To address the challenge of designing throughput-achieving policies under delayed or no channel state information, particularly when channel probing is costly or infeasible.
• To construct tight inner and outer bounds on the capacity region, especially in symmetric, large-user scenarios where the bounds converge.
• To develop a simple, queue-dependent scheduling policy that stabilizes the network for all data rates strictly within the inner capacity bound.
• To provide analytical tools—frame-based Lyapunov drift and stochastic coupling with state aggregation—for analyzing restless bandit problems in this context.

Proposed method

• Derives an inner capacity bound by constructing a queue-dependent scheduling policy that prioritizes users based on queue backlogs and ACK/NACK feedback.
• Uses a frame-based Lyapunov drift argument to prove stability of the proposed policy for all rates within the inner bound, even under heterogeneous traffic.
• Applies stochastic coupling and state aggregation techniques to bound the performance of the restless bandit problem by relating it to a simpler multi-armed bandit system.
• Establishes the tightness of the bounds in the symmetric, large-user regime by showing the inner and outer bounds converge asymptotically.
• Models the channel as a two-state Markov chain (ON/OFF) with memory, leveraging ACK/NACK feedback to infer channel state evolution over time.
• Solves a high-dimensional, partially observed Markov decision problem indirectly by bounding the achievable region rather than solving it exactly.

Experimental results

Research questions

• RQ1 What is the fundamental capacity region of a multi-user wireless downlink when channel states are unknown and modeled as Markov ON/OFF processes?
• RQ2 How can a scheduling policy achieve throughputs close to the theoretical capacity limit under delayed or no instantaneous channel state information?
• RQ3 Under what conditions do the inner and outer bounds on the capacity region become tight?
• RQ4 Can a simple, queue-dependent policy stabilize the network for all data rates within the inner capacity bound?
• RQ5 How can stochastic coupling and state aggregation be used to bound the performance of a restless bandit problem in this wireless network setting?

Key findings

• The proposed queue-dependent scheduling policy stabilizes the network for all data rates strictly within the inner capacity bound, regardless of user heterogeneity or number of users.
• The inner and outer bounds on the capacity region are shown to be tight in the symmetric, large-user regime, indicating the bounds are asymptotically exact.
• The frame-based Lyapunov drift analysis provides a rigorous stability proof for the proposed policy, even under partial and delayed channel state feedback.
• Stochastic coupling and state aggregation techniques successfully bound the performance of the restless bandit problem by relating it to a tractable multi-armed bandit system.
• The results are directly applicable to cognitive radio networks and opportunistic scheduling with uncertain or delayed channel state information.
• The analysis reveals that channel memory, combined with ACK/NACK feedback, can significantly improve network throughput even without explicit channel probing.