HAP-Enabled Communications in Rural Areas: When Diverse Services Meet Inadequate Communication Infrastructures

Yongqiang Zhang, Mustafa A. Kishk, Mohamed-Slim AlouiniIEEE Open J. Commun. Soc 2023 · DOI

Abstract

Bridging the digital divide in rural areas requires innovative network architectures capable of supporting both high-throughput human-type communications (HTC) and freshness-aware machine-type communications (MTC) for applications like smart agriculture. This paper proposes a multi-layer network architecture centered on High Altitude Platforms (HAPs) utilizing Free-Space Optical (FSO) backhaul and Deep Reinforcement Learning (DRL) for dynamic resource allocation. Results demonstrate that the proposed DRL-based approach achieves up to a 2.5x performance improvement over traditional heuristic methods, effectively balancing data throughput and information freshness.

Introduction

Approximately 34% of the global population resides in rural or remote regions lacking adequate internet coverage. For traditional telecommunication operators, the high cost of deploying fiber and terrestrial towers in sparsely populated areas often makes infrastructure expansion economically unfeasible. However, these regions are increasingly in need of modern digital services, ranging from standard web browsing to "freshness-aware" intelligent sensing for smart agriculture and forestry.

The primary challenge lies in heterogeneous resource allocation. While standard internet services (HTC) prioritize high data rates, sensing services (MTC) prioritize the Age of Information (AoI)—a metric quantifying data "freshness." Given the limited infrastructure in rural settings, a single HAP-enabled network must intelligently slice its resources to satisfy these competing requirements.

HAP-enabled multi-layer network architecture for rural communications.

Methods and System Architecture

The researchers investigate a multi-layer network where a HAP, positioned at approximately 20 km altitude, serves as a central hub. The study evaluates three primary architectural variants:

  1. Direct HAP: Devices connect directly to the HAP.
  2. TBS-HAP: Terrestrial Base Stations (TBS) act as relays.
  3. LAP-HAP: Low Altitude Platforms (LAPs), such as UAVs, act as relays to minimize path loss.

A key innovation is the use of Free-Space Optical (FSO) backhaul for the link between relays and the HAP. FSO provides significant bandwidth without licensing constraints, though it remains sensitive to atmospheric conditions like fog.

Dynamic Resource Allocation

To manage the limited communication resources, the authors employ Proximal Policy Optimization (PPO), a DRL algorithm. Unlike static time-slicing, the DRL agent dynamically adjusts resource allocation ratios based on real-time network states, including channel gains, sensor AoI, and required uplink rates. The study compares Identical Dynamic Scheduling (IDS) with Non-identical Dynamic Scheduling (NIDS), finding that NIDS allows for more granular, cell-specific control.

Timing diagram of resource allocation between HTC and MTC small-cells.

Results and Key Insights

The simulation results highlight several critical findings:

  • Superiority of DRL: The DRL-based dynamic method outperformed the Ant Colony Optimization (ACO) heuristic by nearly 2.5 times in weighted sum performance.
  • Architecture Efficiency: The LAP-FSO architecture (UAV relays with optical backhaul) proved most effective. Compared to a HAP-only setup, the LAP architecture with RF backhaul improved AoI by 22.5% and data rates by 12.9%.
  • Non-Linearity of Freshness: While data rates scale linearly with resources, AoI does not. The DRL framework successfully "over-allocates" resources to sensors at critical moments to maintain freshness, achieving an average AoI of 2.6 seconds alongside an 11 Mbps data rate.
  • FSO as a Capacity Booster: Despite weather vulnerabilities, FSO backhaul significantly increased capacity, enabling the network to meet HTC demands without compromising MTC sensor freshness.

Conclusion and Future Work

This research demonstrates that HAP-based multi-layer networks, powered by DRL and FSO, offer a viable solution for rural connectivity. However, challenges remain regarding FSO weather sensitivity and the energy constraints of untethered UAVs. Future work will focus on hybrid RF/FSO systems, distributed multi-agent RL for improved scalability, and the optimization of aerial base station placement to further enhance network robustness and coverage.

Summary updated

BibTeX
@article{zhang2023hap,
  author = {Yongqiang Zhang and Mustafa A. Kishk and Mohamed-Slim Alouini},
  title = {{HAP-Enabled Communications in Rural Areas: When Diverse Services Meet Inadequate Communication Infrastructures}},
  journal = {IEEE Open J. Commun. Soc.},
  year = {2023},
  doi = {10.1109/ojcoms.2023.3318836},
}
Figure