On-Time Communications Over Fading Channels

Yan Li, Yunquan Dong, Jian Wang, Byonghyo Shim

TL;DR

This paper tackles the challenge of ensuring "on-time" packet delivery in wireless systems, which is crucial for emerging applications like industrial automation and autonomous driving. The authors propose a novel metric, the "δ-on-time reception rate," and develop an optimal packet scheduling policy using Markov Decision Processes (MDPs) to significantly improve this rate compared to random transmissions.

What Problem Does This Paper Address?

Modern industrial and communication systems, such as the Industrial Internet of Things (IIoT), 5G URLLC, and autonomous driving, demand highly reliable and timely data delivery. This means not just low latency, but also predictable delivery within a tight time window. Traditional "best-effort" communication methods are insufficient for these stringent requirements. Wireless channels, inherently prone to fading and unpredictable variations, make guaranteeing such precise delivery even more challenging. The paper aims to quantify and improve the "on-time" reception of packets over these unreliable wireless links, moving beyond simple latency measurements to consider a defined reception window.

Methods

The core of the paper lies in two main areas:

  • Defining and Analyzing On-Time Reception: The authors introduce the concept of "δ-on-time reception," where a packet is considered on-time if it arrives within a specified window around its target delivery slot (target slot ± δ slots). They then derive analytical expressions for the probability of a single packet being δ-on-time and the overall δ-on-time reception rate for random transmissions over a fading channel.
  • Optimal Packet Scheduling via MDP: To improve upon random transmissions, the paper formulates the packet scheduling problem as a Markov Decision Process (MDP). The system state is defined by the difference between the current time and the target reception time. The available actions include delaying, dropping, or repeating packets. An efficient iterative algorithm is used to find the optimal scheduling policy that maximizes the expected on-time reception rate.

Datasets

Not discussed in the paper. The paper relies on analytical derivations and simulations with synthetic channel models.

Key Insights and Ideas

  • Beyond Latency: The δ-on-time Metric: Simply measuring latency isn't enough for time-sensitive applications. The δ-on-time metric provides a more practical measure of timeliness by accounting for a small, acceptable deviation.
  • Random Transmissions Degrade with Sequence Length: The on-time reception rate for random transmissions significantly decreases as the number of packets in a sequence increases, eventually approaching zero. This highlights the inadequacy of random approaches for sustained on-time delivery.
  • Optimal Scheduling is Crucial: The proposed MDP-based optimal scheduling policy dramatically improves the on-time reception rate, allowing it to converge to a much larger constant value even for long packet sequences.
  • Repeating is the Dominant Strategy: Simulations show that the "repeating" strategy is the most frequently used action under the optimal policy, suggesting that retransmitting packets is key to achieving on-time delivery in fading channels.
  • Trade-offs Exist: While optimal scheduling significantly boosts on-time reception, it cannot completely eliminate the impact of channel randomness. The paper demonstrates that even with optimal policies, 100% on-time delivery is not achievable.

Key Results

  • Analytical Framework: The paper provides a solid analytical framework for evaluating on-time communication performance, including the derivation of probabilities for δ-on-time reception.
  • Random Transmission Performance: The on-time reception rate for random transmissions was shown to decrease monotonically with the sequence length, approaching zero.
  • Optimal Scheduling Improvement: The optimal packet scheduling policy, derived via MDP, significantly outperforms random transmissions. For instance, in Figure 7, the optimal policy achieves an on-time reception rate of approximately 0.7 compared to around 0.3 for random transmissions when δ=2. Figure 8 shows that even for strictly on-time delivery (δ=0), the optimal policy increases the rate by about 13%.
  • Convergence: The on-time reception rate with optimal scheduling converges to a stable, higher constant as the number of packets increases (Figure 10).
  • Strategy Proportions: In a specific simulation scenario (Figure 11), the optimal policy heavily favors repeating packets (61.9%), followed by delaying (36.3%), with dropping being minimal (1.6%).

Limitations and Future Work

  • Channel Model: The analysis primarily relies on a Rayleigh fading channel model. While common, real-world wireless environments can be more complex.
  • State Space: The MDP formulation can lead to a large state space. The paper addresses this by limiting the state space to a finite range, which is a practical approximation.
  • Future Work: The authors suggest further research into power allocation, variable-rate compressions, and parallel transmissions to further combat fading channel randomness and improve on-time reception. They also aim to jointly optimize the δ-on-time reception rate for TSN networks.

Summary updated

Abstract

We consider the on-time transmissions of a sequence of packets over a fading channel. Different from traditional in-time communications, we investigate how many packets can be received <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\delta$</tex-math></inline-formula>-on-time, meaning that the packet is received around an expected slot with a deviation no larger than <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\delta$</tex-math></inline-formula> slots. In this framework, we first derive the on-time reception rate of the random transmission scheme when no packet-controlling is used. Our analytical and simulation results show that the on-time reception rate of random transmissions decreases (to zero) with the sequence length. To improve the on-time reception rate, we further propose to schedule the packets by delaying, dropping, or repeating the transmissions. Specifically, we model the packet scheduling problem as a Markov decision process (MDP) and then obtain the optimal scheduling policy using an efficient iterative algorithm. By using the optimal packet scheduling, the on-time reception rate converges to a much larger constant, thus ensuring better on-timeliness. Moreover, we show that the on-time reception rate increases if the target reception interval and/or the deviation tolerance <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\delta$</tex-math></inline-formula> is increased, or the randomness of the fading channel is reduced. By optimally scheduling the packets and choosing these parameters, therefore, wireless channels will be more suitable for the latency-sensitive applications.
BibTeX
@article{li2022time,
  author = {Yan Li and Yunquan Dong and Jian Wang and Byonghyo Shim},
  title = {{On-Time Communications Over Fading Channels}},
  journal = {IEEE Trans. Vehicular Technology},
  year = {2022},
  doi = {10.1109/tvt.2023.3303570},
}
Figure