Abstract
Space-air-ground integrated networks (SAGINs) will play a key role in 6G communication systems. They are considered a promising technology to enhance the network capacity in highly dense agglomerations and to provide connectivity in rural areas. The multi-layer and heterogeneous nature of SAGINs necessitates an innovative design of their multi-tier associations. We propose a modeling of the SAGINs association problem using multi-sided matching theory. Our aim is to provide a reliable, asynchronous and fully distributed approach that associates nodes across the layers so that the total end-to-end rate of the assigned agents is maximized. To this end, our problem is modeled as a multi-sided many-to-one matching game. A randomized matching algorithm with low information exchange is proposed. The algorithm is shown to reach an efficient and stable association between nodes in adjacent layers. Our simulation results show that the proposed approach achieves significant gain compared to the greedy and distance-based algorithms.
The Challenge of SAGIN Association
As we move toward 6G, Space-Air-Ground Integrated Networks (SAGINs) are emerging as a critical solution for global coverage, reaching remote areas and providing resilient connectivity during disasters. These networks are inherently complex, comprising multiple tiers: satellites in space, High-Altitude Platforms (HAPs) and Unmanned Aerial Vehicles (UAVs) in the air, and various devices on the ground.
The core technical challenge lies in association: determining the optimal path for a ground user through a UAV, then a HAP, and finally to a satellite. Traditional resource allocation methods are typically centralized, requiring a single controller to maintain global state knowledge. This approach suffers from massive communication overhead, poor scalability, and slow response times in the face of high-speed aerial and space node dynamics. This research proposes a distributed, asynchronous framework where nodes autonomously find optimal partners.
Decentralized Matching Methodology
The SAGIN association problem is modeled as a multi-sided many-to-one matching game with Transferable Utility (TU). This framework treats the network as a market where nodes in different layers "trade" to form high-value communication chains.
The proposed Blind Matching Algorithm (BLMA) operates through several key mechanisms:
- Aspiration Learning: Nodes initialize with a target data rate (aspiration level).
- Random Encounters: Nodes randomly interact with potential partners in adjacent layers.
- Local Decisions: Matches are formed if both nodes can improve their utility by at least a small margin ($\epsilon$).
- Blindness: Nodes require no global network knowledge, only observing their own potential payoff from a specific partner.
- Stability: Nodes that remain unassociated gradually lower their aspiration levels (by $\delta$) until the system reaches an $\epsilon$-pairwise stable state.
A critical insight is that because the total value of a communication chain is the sum of its individual links (local additivity), a stable solution is guaranteed to exist.
Performance Evaluation
The BLMA was evaluated in a simulated suburban environment with 40 ground devices, 8 UAVs, 3 HAPs, and 2 satellites. It was compared against two baselines: a centralized Greedy Algorithm and a Distance-based (Closest Association) Algorithm.
Key findings include:
- Throughput Gains: For 40 users, BLMA achieved an end-to-end rate of approximately 190 Mbits/s, significantly outperforming the Greedy (~82 Mbits/s) and Distance-based (~48 Mbits/s) approaches.
- Scalability: The performance gap between BLMA and simpler baselines widens as user density increases, making it ideal for dense 6G deployments.
- Efficiency: Despite its iterative nature, the algorithm converges rapidly. For a network of 300 users, average convergence time was less than 1.4 seconds.
Conclusion and Future Directions
This work demonstrates that a decentralized, randomized matching approach can effectively manage the complex associations in multi-layered SAGINs with minimal overhead. While the current model assumes static snapshots, future research will address high-speed mobility (e.g., LEO satellites), bi-directional traffic, and explicit energy constraints for battery-powered nodes like UAVs.
Summary updated