TL;DR
This paper introduces a Unified Cellular Scheduling (UCS) framework designed to ensure predictable communication reliability in cellular networks that incorporate Device-to-Device (D2D) links. By unifying the scheduling of cellular and D2D transmissions and leveraging a novel interference model, UCS aims to maximize spatial reuse while guaranteeing reliability for mission-critical applications.
What Problem Does This Paper Address?
Modern cellular networks, particularly with the advent of LTE-Advanced Pro and 5G, are increasingly supporting Device-to-Device (D2D) communications for mission-critical applications like real-time control and augmented reality (AR/VR). A key challenge for these applications is predictable communication reliability. Unpredictable reliability can lead to delays and system-level instability. A major source of this unpredictability in cellular networks is interference between concurrent transmissions, especially when D2D links are introduced alongside traditional cellular uplinks and downlinks. Existing solutions often struggle to manage this interference effectively across different link types and multiple communication channels, leading to suboptimal performance or an inability to guarantee reliability.
Methods
The core of UCS is its Unified Cellular Scheduling (UCS) framework. This framework is built upon the Physical-Ratio-K (PRK) interference model, which is designed to be high-fidelity and suitable for distributed scheduling.
Key methodological contributions include:
- Unified Scheduling: UCS schedules uplink, downlink, and D2D transmissions in a single, unified manner. This contrasts with approaches that might treat these link types separately.
- PRK Interference Model: This model is adapted to account for inter-cell interference and the complexities of multi-channel environments. It allows for the instantiation of a parameter 'K' that defines an "exclusion region" around a receiver, indicating which potential interferers are too strong to transmit concurrently without compromising reliability.
- Mode Selection: For UE-to-UE communication pairs, UCS employs a mode selection mechanism that decides whether to use a cellular mode (via the base station) or a direct D2D mode. This decision is based on minimizing the interference impact, modeled as a Restless Multi-Armed Bandit (MAB) problem.
- Multi-Channel Operation: UCS extends PRK-based scheduling to multi-channel settings. It groups adjacent carriers under a single PRK parameter 'K' to reduce control signaling overhead while still enabling channel hopping for resilience against fading and interference.
- BS-UE Orchestration: UCS leverages the existing cellular infrastructure by placing most intelligence at the Base Stations (BSes) and keeping User Equipment (UEs) functionality minimal. UEs report local state information (like channel gains and reliability estimates) to BSes, which then orchestrate scheduling decisions. This facilitates incremental deployment and technology evolution.
- ONAMA Scheduling Extension: The core scheduling algorithm is an extension of the ONAMA (Optimal Maximal Concurrency) TDMA algorithm, adapted for multi-channel cellular networks with traffic demand awareness.
Datasets
Not discussed in the paper. The evaluation relies on simulations and a small-scale hardware testbed.
Key Insights and Ideas
- Unified Interference Management is Crucial: Treating cellular and D2D links as distinct entities leads to suboptimal interference management. A unified approach, like UCS, is necessary for predictable reliability.
- PRK Model's Adaptability: The PRK model, when properly instantiated and adapted, provides a robust foundation for predictable interference control across diverse link types and network conditions.
- Trade-offs in Multi-Channel Scheduling: Grouping carriers under a single PRK parameter 'K' is essential for managing control overhead in multi-channel cellular networks. However, there's a trade-off: too many carriers per 'K' can reduce accuracy, while too few increases overhead. UCS finds a balance.
- BS-Centric Intelligence: Offloading scheduling intelligence to BSes, while keeping UEs lightweight, is a practical approach for incremental deployment and integration with existing cellular standards.
- Channel Hopping for Resilience: The multi-channel nature of UCS, combined with its scheduling, naturally enables channel hopping, which significantly improves resilience to fading and external interference.
Key Results
- Predictable Reliability Guaranteed: UCS consistently ensures the application-required packet-delivery-reliability (PDR) across various scenarios, outperforming existing schemes like IAS and QAS, which fail to guarantee high reliability (e.g., 95%).
- High Spatial Reuse: UCS achieves a channel spatial reuse rate statistically equal to the state-of-the-art centralized algorithm iOrder, demonstrating its efficiency in maximizing concurrent transmissions.
- Reduced Exclusion Region: By dynamically adjusting the PRK parameter 'K' based on bandwidth and reliability requirements, UCS effectively reduces the size of exclusion regions, allowing for more concurrent transmissions.
- Mode Selection Effectiveness: The mode selection mechanism correctly identifies D2D as the preferred mode for UE-to-UE communication pairs when it offers a smaller interference footprint.
- Hardware Validation: A small-scale deployment using USRP B210 SDRs and OpenAirInterface successfully validated the feasibility of the UCS framework.
Limitations and Future Work
- Static/Slowly Moving UEs: The current work assumes UEs are static or move slowly, simplifying channel gain and background noise stability. The authors note that for highly mobile networks, techniques for dynamic power control and mobility management would be needed, but a detailed study is beyond the scope.
- Parameter 'K' Selection: While the paper discusses the trade-offs, finding the optimal number of carriers to share a single PRK parameter 'K' for all scenarios might require further investigation.
- Future Work: The authors suggest that techniques for handling highly mobile networks and power control could be integrated. Further exploration of the trade-offs in multi-channel scheduling and the impact of different channel models and network topologies is also implied by the extensive simulation setup.
Summary updated