Introduction
As we transition toward the 6G era, the demand for ubiquitous, high-speed, and low-latency computational resources is reaching an inflection point. While edge and cloud computing have significantly advanced data processing capabilities, they remain fundamentally tethered to terrestrial infrastructure. This reliance creates critical gaps in coverage—particularly in rural areas, oceans, and disaster zones—and struggles to support high-mobility entities such as autonomous vehicles and "flying taxis."
Aerial computing emerges as a transformative 6G paradigm that merges Aerial Access Networks (AANs) with edge computing to create a multi-tier, "in-the-sky" processing infrastructure. By integrating Unmanned Aerial Vehicles (UAVs), high-altitude platforms, and Low-Earth-Orbit (LEO) satellites, this framework provides global computational resources for applications that terrestrial networks cannot reach.
The Aerial Computing Architecture
The proposed framework is defined by a hierarchical, four-layer architecture designed to optimize data flow and processing efficiency across varying altitudes and coverage areas.
- IoT Layer: Ground-level sensors and devices that generate raw data from vertical domains like smart cities and factories.
- Low-Altitude Computing (LAC): Drones and UAVs operating at 0–10 km. These nodes provide rapid, cost-effective deployment for localized, time-sensitive tasks.
- High-Altitude Computing (HAC): Airships and balloons positioned at 17–50 km. These offer wider coverage (up to 10 km cell size) and higher endurance, acting as quasi-stationary aerial servers.
- Satellite Computing: LEO satellites at 80–2000 km. This layer provides a "global umbrella," ensuring connectivity and compute relay for sparse users across the planet.
A New Paradigm: Comparison with Existing Technologies
Aerial computing is not merely an extension of existing edge technologies but a distinct paradigm. Unlike Cloudlets, Fog Computing, or Multi-access Edge Computing (MEC), which are largely localized or terrestrial-bound, aerial computing offers global service coverage and "very high" mobility and scalability.
As shown in the comparison above, aerial computing introduces a "medium" latency profile—slightly higher than local MEC but significantly more flexible. Its architecture is uniquely localized, hierarchical, and on-demand, making it suitable for sparse user distributions where traditional infrastructure is non-viable.
Key Insights and Enabling Technologies
The core innovation of aerial computing lies in the Amalgamation Factor: transforming aerial nodes from simple communication relays (pipes) into active processing servers (brains). This shift is supported by several key-enabling technologies:
- Network Softwarization: Utilizing Software-Defined Networking (SDN) and Network Function Virtualization (NFV) to manage the dynamic, 3D topology of aerial nodes.
- Hierarchical Offloading: Tasks are intelligently distributed across the stack. A ground sensor might offload data to a drone (LAC), which then offloads heavier sub-tasks to an airship (HAC) or satellite, optimizing for energy and latency at every step.
- AI Integration: Artificial Intelligence serves a dual role. Machine learning optimizes network trajectories and resource allocation, while the network itself acts as a distributed platform for Federated Learning (FL), allowing data to be processed locally on aerial nodes to preserve privacy.
Quantitative Gains and Applications
Research synthesized in this survey demonstrates significant performance improvements across various sectors:
- Smart Cities: 3D placement schemes for aerial nodes have improved information coverage by 21.42% while reducing cost ratios by 34.32%.
- Industrial Monitoring: Deep learning models on LAC servers achieved 98.38% accuracy in detecting power line insulator defects and 93.58% accuracy in crop row detection.
- Energy Efficiency: HAC-borne edge intelligence has been shown to generate 44% more power for solar-tracking UAVs, reducing flight time requirements by 25%.
- Global Latency: LEO satellite systems currently provide latencies between 20–40 ms, making them competitive with terrestrial mobile networks for remote compute tasks.
Challenges and Future Directions
Despite its promise, several "hard problems" remain for the widespread adoption of aerial computing:
- The Energy Bottleneck: Sustainable long-term operation remains a challenge due to the limited battery life of drones, necessitating further research into solar power and Wireless Power Transfer (WPT).
- Dynamic Stability: Maintaining a stable compute fabric with nodes flying at different speeds and altitudes requires highly complex handover protocols.
- Security: Data transmitted through the air is inherently vulnerable to interception. Future research must prioritize Blockchain integration and Physical Layer Security (PLS) to protect the aerial fabric.
In conclusion, aerial computing represents a vital step toward the 6G vision of a truly connected, intelligent world, bridging the gap between terrestrial limitations and global computational needs.
Summary updated