Dynamic Power Control for Delay-Aware Device-to-Device Communications

Zhang Fan (Michael), Wei Wang, Fan Zhang, Vincent K. N. LauIEEE JSAC 2015 · DOI

TL;DR

This paper tackles the challenge of optimizing power control in device-to-device (D2D) communication systems, aiming to minimize both average transmit power and data flow delays. They achieve this by formulating the problem as a Markov Decision Process and developing a low-complexity, delay-aware power control algorithm that leverages interference filtering properties of the MAC protocol and a novel approximation of the priority function.

What Problem Does This Paper Address?

Future wireless networks, like IMT-advanced, are expected to support higher data rates and capacities. Device-to-device (D2D) communication, where devices communicate directly, is a promising technology to achieve this by offering high data rates, low delays, and reduced power consumption. However, D2D communication introduces new challenges, particularly in managing interference between D2D pairs and cellular users, and in meeting the stringent delay requirements of modern applications like video streaming and VoIP. Existing power control methods often focus solely on physical layer throughput or ignore the bursty nature of data arrivals and the critical need for low latency. This paper addresses the need for a dynamic power control strategy that explicitly considers both the instantaneous channel conditions and the urgency of data queues to optimize for delay-aware D2D communications.

Methods

The authors formulate the problem as an infinite horizon average cost Markov Decision Process (MDP). To overcome the "curse of dimensionality" inherent in K-dimensional MDPs (where K is the number of D2D pairs), they exploit two key properties:

  1. CSMA-like MAC Protocol: This protocol inherently filters interference, leading to a "weak coupling" property among D2D nodes.
  2. Base Station (BS) Assistance: The BS simplifies control information exchange.

They derive a simplified optimality condition for the MDP, which involves solving a K-dimensional partial differential equation (PDE). To make this tractable, they use perturbation analysis on the interference filtering property of the MAC protocol to derive a closed-form approximate priority function and its error bound. This approximate priority function is then used to develop a low-complexity, per-stage optimization algorithm for power control. The proposed solution is shown to be asymptotically optimal for a sufficiently large carrier sensing distance.

Datasets

Not discussed in the paper. The evaluation is based on simulations.

Key Insights and Ideas

  • Delay is Crucial: The paper emphasizes that for many real-world applications, minimizing delay is as important, if not more important, than maximizing throughput.
  • Interference Filtering is Key: The CSMA-like MAC protocol, often seen as a source of contention, is cleverly leveraged here to reduce the complexity of the optimization problem by weakening the coupling between D2D queues.
  • Approximation is Powerful: The core innovation lies in approximating the complex priority function (which dictates queue urgency) with a closed-form solution. This bypasses the need for computationally expensive exact solutions.
  • Cross-Layer Design Pays Off: By jointly considering physical layer (channel state information - CSI) and higher layers (queue state information - QSI), the algorithm can make more informed and effective power control decisions.
  • Asymptotic Optimality: The proposed algorithm is not just a heuristic; it's proven to converge to the optimal solution under certain conditions (large carrier sensing distance), providing theoretical guarantees.

Key Results

The paper demonstrates significant performance gains through simulations. Compared to various baseline schemes (Cellular Mode, Fixed Power D2D, CSI-Based Power Control, Queue-Weighted Power Control), the proposed dynamic power control scheme consistently achieves lower average delays across different scenarios, including varying average arrival rates, transmit powers, and D2D communication ranges. For instance, Figure 2 shows that the proposed scheme significantly outperforms baselines, especially under high traffic loads. Figure 5 illustrates that the proposed scheme maintains good delay performance over a wide range of carrier sensing distances. Table II also shows that the computational complexity of the proposed scheme is comparable to other advanced methods but significantly lower than brute-force value iteration, making it practical for real-time deployment.

Limitations and Future Work

  • Carrier Sensing Distance: The performance of the proposed scheme is sensitive to the carrier sensing distance. While it achieves good performance over a large regime, very small distances can lead to inefficient interference filtering, and very large distances can reduce spatial reuse. The paper suggests selecting an appropriate distance to balance these tradeoffs.
  • Asymptotic Optimality: The asymptotic optimality is proven for sufficiently large carrier sensing distances. The performance for smaller distances might deviate from the optimal.
  • Future Work: The authors mention that the solution framework can be extended to multi-channel systems (like OFDMA) and general fading distributions. They also suggest further exploration of the trade-off between delay and other performance metrics.

Summary updated

BibTeX
@article{michael2015dynamic,
  author = {ZHANG Fan (Michael) and Wei Wang and Fan Zhang and Vincent K. N. Lau},
  title = {{Dynamic Power Control for Delay-Aware Device-to-Device Communications}},
  journal = {IEEE JSAC},
  year = {2015},
  doi = {10.1109/jsac.2014.2369614},
}
Figure