[Paper Review] Dynamic Power Control for Time-Critical Networking with Heterogeneous Traffic

Emmanouil Fountoulakis, Νικόλαος Παππάς+1 more|arXiv (Cornell University)|Nov 9, 2020

Advanced Wireless Network Optimization35 references4 citations

TL;DR

This paper proposes the Dynamic Power Control (DPC) algorithm, a Lyapunov optimization-based real-time scheduling scheme for time-critical wireless networks with heterogeneous traffic. It minimizes packet drop rate while guaranteeing minimum throughput and respecting user power constraints, achieving a solution arbitrarily close to optimal with strong stability guarantees under time-varying channels and mobility.

ABSTRACT

Future wireless networks will be characterized by heterogeneous traffic requirements. Such requirements can be low-latency or minimum-throughput. Therefore, the network has to adjust to different needs. Usually, users with low-latency requirements have to deliver their demand within a specific time frame, i.e., before a deadline, and they co-exist with throughput oriented users. In addition, the users are mobile and they share the same wireless channel. Therefore, they have to adjust their power transmission to achieve reliable communication. However, due to the limited power budget of wireless mobile devices, a power-efficient scheduling scheme is required by the network. In this work, we cast a stochastic network optimization problem for minimizing the packet drop rate while guaranteeing a minimum throughput and taking into account the limited-power capabilities of the users. We apply tools from Lyapunov optimization theory in order to provide an algorithm, named Dynamic Power Control (DPC) algorithm, that solves the formulated problem in realtime. It is proved that the DPC algorithm gives a solution arbitrarily close to the optimal one. Simulation results show that our algorithm outperforms the baseline Largest-Debt-First (LDF) algorithm for short deadlines and multiple users.

Motivation & Objective

• To address the challenge of scheduling deadline-constrained and throughput-oriented users in time-critical wireless networks with limited power budgets.
• To design a real-time, power-efficient scheduling algorithm that balances low-latency and minimum-throughput requirements under dynamic channel conditions.
• To guarantee strong stability of queues and minimize packet drop rates while respecting user power limitations.
• To provide a solution that is arbitrarily close to the optimal through Lyapunov optimization theory.

Proposed method

• Formulates a stochastic network optimization problem to minimize packet drop rate under minimum-throughput and power budget constraints.
• Applies Lyapunov optimization to derive a dynamic power control (DPC) algorithm that makes real-time decisions based on queue backlogs and channel state information.
• Uses a drift-plus-penalty framework to balance queue stability and performance, with a penalty term representing the drop rate.
• Introduces virtual queues for throughput and deadline constraints to enforce long-term average performance guarantees.
• Derives a policy that selects power allocations to minimize a combination of queue backlog and penalty terms at each time slot.
• Proves that the DPC algorithm achieves a solution arbitrarily close to the optimal by tuning the Lyapunov parameter V.

Experimental results

Research questions

• RQ1 How can a real-time scheduling algorithm be designed to minimize packet drop rates in time-critical wireless networks with heterogeneous traffic?
• RQ2 What is the optimal trade-off between minimizing drop rate, guaranteeing minimum throughput, and respecting user power constraints?
• RQ3 Can a dynamic power control algorithm achieve near-optimal performance while ensuring strong stability of queues under time-varying channels?
• RQ4 How does the proposed DPC algorithm compare to baseline schemes like Largest-Debt-First (LDF) in terms of performance under short deadlines and multiple users?
• RQ5 What theoretical guarantees can be provided for the stability and performance of the dynamic scheduling policy?

Key findings

• The DPC algorithm achieves a solution arbitrarily close to the optimal by tuning the Lyapunov parameter V, with a performance gap bounded by O(1/V).
• The algorithm ensures strong stability of all queues, meaning the time-averaged queue sizes remain bounded, which implies that constraints on throughput and deadline compliance are satisfied almost surely.
• Simulation results show that DPC outperforms the baseline Largest-Debt-First (LDF) algorithm in terms of reducing packet drop rate, especially under short deadlines and high user loads.
• The theoretical analysis proves that the time-averaged drop rate is bounded within O(1/V) of the optimal, and the time-averaged power usage remains within the user's power budget.
• The algorithm maintains stability even under high mobility and time-varying channel conditions by dynamically adapting power based on real-time queue and channel state information.
• The DPC algorithm provides a scalable solution that avoids the curse of dimensionality, unlike Markov decision process-based approaches, making it suitable for larger network scenarios.