Unified Scheduling for Predictable Communication Reliability in Cellular Networks with D2D Links

Yuwei Xie, Hongwei Zhang, Pengfei Ren

Abstract

Cellular networks with D2D links are increasingly being explored for mission-critical applications (e.g., real-time control and AR/VR) which require predictable communication reliability. Thus it is critical to control interference among concurrent transmissions in a predictable manner to ensure the required communication reliability. To this end, we propose a Unified Cellular Scheduling (UCS) framework that, based on the Physical-Ratio-K (PRK) interference model, schedules uplink, downlink, and D2D transmissions in a unified manner to ensure predictable communication reliability while maximizing channel spatial reuse. UCS also provides a simple, effective approach to mode selection that maximizes the communication capacity for each involved communication pair. UCS effectively uses multiple channels for high throughput as well as resilience to channel fading and external interference. Leveraging the availability of base stations (BSes) as well as high-speed, out-of-band connectivity between BSes, UCS effectively orchestrates the functionalities of BSes and user equipment (UE) for light-weight control signaling and ease of incremental deployment and integration with existing cellular standards. We have implemented UCS using the open-source, standards-compliant cellular networking platform OpenAirInterface. We have validated the OpenAirInterface implementation using USRP B210 software-defined radios and lab deployment. We have also evaluated UCS through high-fidelity, at-scale simulation studies; we observe that UCS ensures predictable communication reliability while achieving a higher channel spatial reuse rate than existing mechanisms, and that the distributed UCS framework enables a channel spatial reuse rate statistically equal to that in the state-of-the-art centralized scheduling algorithm iOrder.

Introduction

As cellular networks evolve toward 5G and beyond, their utility extends far beyond traditional mobile broadband. Emerging "mission-critical" applications—such as remote surgery, autonomous vehicle coordination, and industrial automation—demand a level of reliability that current "best-effort" mechanisms struggle to provide. Specifically, these applications require predictable reliability, where packet delivery must meet a high, guaranteed probability (e.g., 95% or 99%) within strict latency bounds.

A significant challenge in modern cellular environments is the integration of Device-to-Device (D2D) links. While D2D communication can significantly improve spectral efficiency and reduce latency by allowing direct communication between nearby devices, it introduces a complex interference landscape. Traditional LTE interference management, such as the High Interference Indicator (HII), is often reactive and lacks the precision to guarantee specific reliability targets. Furthermore, existing research often relies on oversimplified interference models (like the protocol/disk model) that fail to capture the nuances of real-world signal propagation and inter-cell interference.

The Unified Cellular Scheduling (UCS) Framework

To address these challenges, we propose the Unified Cellular Scheduling (UCS) framework. UCS is designed to manage uplink, downlink, and D2D transmissions within a single, cohesive scheduling logic. The framework is built upon three core technical pillars.

1. The Physical-Ratio-K (PRK) Interference Model

At the heart of UCS is the PRK model, which provides a more accurate representation of interference than traditional distance-based models. In the PRK model, a receiver $R$ defines an "exclusion region" around itself. Any other concurrent transmitter $C$ must be far enough away such that the ratio of the signal power from its intended transmitter $S$ to the interference power from $C$ exceeds a certain threshold.

The PRK model defines an exclusion region around a receiver to manage interference and ensure reliability.

The innovation in UCS is treating the PRK parameter $K$ as a control variable. By monitoring the real-time Packet Delivery Ratio (PDR), UCS can dynamically adjust the size of the exclusion region. If the PDR falls below the target, the exclusion region expands (increasing $K$) to reduce interference. Conversely, if the PDR exceeds the target, the region can shrink to allow more concurrent transmissions, thereby maximizing spatial reuse.

2. Multi-Armed Bandit (MAB) Mode Selection

When two devices need to communicate, they can choose between D2D mode (direct) or Cellular mode (via the Base Station). UCS treats this choice as a learning problem. Using a multi-armed bandit approach—specifically the HD-SEE algorithm—the system balances "exploration" (trying different modes to see their current performance) and "exploitation" (using the mode that currently offers the best reliability-to-capacity tradeoff). This allows the network to adapt to changing channel conditions and device locations automatically.

3. Multi-Channel ONAMA Scheduling

UCS extends the distributed ONAMA (Optimized Network-Assisted Medium Access) scheduling protocol to operate across multiple frequency carriers. This multi-channel approach provides frequency diversity, making the system more resilient to narrow-band fading and external interference. By leveraging the high-speed backhaul (X2 interface) between Base Stations (BSes), UCS achieves a distributed scheduling architecture that avoids the latency and single-point-of-failure risks of a centralized controller.

The UCS architecture orchestrates functions between the Base Station (BS) and User Equipment (UE) across control and data planes.

Implementation and Evaluation

We implemented UCS on the OpenAirInterface (OAI) platform, a standards-compliant open-source cellular stack. The implementation was validated using USRP B210 software-defined radios in a real-world laboratory environment.

To complement the hardware tests, we conducted extensive high-fidelity simulations using the OAI simulator. The simulation setup involved 135 User Equipments (UEs) distributed across 9 cells. We tested various network topologies (random and grid) and channel models, including Rayleigh, Rice, and AWGN fading, with different path-loss exponents.

Key Results

The evaluation demonstrated that UCS consistently meets its design goals:

  • Predictable Reliability: In both hardware and simulation, UCS successfully maintained target PDRs (e.g., 90% and 95%) even under heavy network load.
  • Superior Spatial Reuse: UCS outperformed existing interference-aware (IAS) and QoS-aware (QAS) protocols. While IAS often failed to maintain reliability above 60% in high-interference scenarios, UCS remained at its 95% target.
  • Scalability and Efficiency: By grouping carriers and sharing PRK parameters, UCS reduced control signaling overhead by over 90% compared to per-carrier management, without any loss in reliability. Furthermore, the distributed UCS framework achieved spatial reuse rates statistically equivalent to the state-of-the-art centralized scheduler, iOrder.

Conclusion and Future Work

UCS represents a significant step toward enabling mission-critical applications in cellular networks. By unifying the management of all link types under a control-theoretic interference model, it provides the predictable reliability required for the next generation of industrial and consumer services. Future work will focus on extending UCS to support high-mobility scenarios, integrating dynamic power control, and optimizing performance for bursty IoT traffic patterns.

Summary updated

Abstract

Cellular networks with D2D links are increasingly being explored for mission-critical applications (e.g., real-time control and AR/VR) which require predictable communication reliability. Thus it is critical to control interference among concurrent transmissions in a predictable manner to ensure the required communication reliability. To this end, we propose a Unified Cellular Scheduling (UCS) framework that, based on the Physical-Ratio-K (PRK) interference model, schedules uplink, downlink, and D2D transmissions in a unified manner to ensure predictable communication reliability while maximizing channel spatial reuse. UCS also provides a simple, effective approach to mode selection that maximizes the communication capacity for each involved communication pair. UCS effectively uses multiple channels for high throughput as well as resilience to channel fading and external interference. Leveraging the availability of base stations (BSes) as well as high-speed, out-of-band connectivity between BSes, UCS effectively orchestrates the functionalities of BSes and user equipment (UE) for light-weight control signaling and ease of incremental deployment and integration with existing cellular standards. We have implemented UCS using the open-source, standards-compliant cellular networking platform OpenAirInterface. We have validated the OpenAirInterface implementation using USRP B210 software-defined radios and lab deployment. We have also evaluated UCS through high-fidelity, at-scale simulation studies; we observe that UCS ensures predictable communication reliability while achieving a higher channel spatial reuse rate than existing mechanisms, and that the distributed UCS framework enables a channel spatial reuse rate statistically equal to that in the state-of-the-art centralized scheduling algorithm iOrder.
BibTeX
@inproceedings{xie2020unified,
  author = {Yuwei Xie and Hongwei Zhang and Pengfei Ren},
  title = {{Unified Scheduling for Predictable Communication Reliability in Cellular Networks with D2D Links}},
  booktitle = {Computer},
  year = {2020},
  doi = {10.1016/j.comcom.2020.12.012},
}

Loading related content...

Figure