Abstract
The aerial-terrestrial communication system constitutes an efficient paradigm for supporting and complementing terrestrial communications. However, the benefits of such a system cannot be fully exploited, especially when the line-of-sight (LoS) transmissions are prone to severe deterioration due to complex propagation environments in urban areas. The emerging technology of reconfigurable intelligent surfaces (RISs) has recently become a potential solution to mitigate propagation-induced impairments and improve wireless network coverage. Motivated by these considerations, in this paper, we address the coverage and link performance problems of the aerial-terrestrial communication system by proposing an RIS-assisted transmission strategy. In particular, we design an adaptive RIS-assisted transmission protocol, in which the channel estimation, transmission strategy, and data transmission are independently implemented in a frame. On this basis, we formulate an RIS-assisted transmission strategy optimization problem as a mixed-integer non-linear program (MINLP) to maximize the overall system throughput. We then employ multi-task learning to speed up the solution to the problem. Benefiting from multi-task learning, the computation time is reduced by about four orders of magnitude. Numerical results show that the proposed RIS-assisted transmission protocol significantly improves the system throughput and reduces the transmit power.
Introduction
Aerial-terrestrial communication systems, utilizing Unmanned Aerial Vehicles (UAVs) as flying base stations, offer a flexible solution for extending wireless coverage and supporting terrestrial networks. However, in dense urban environments, the critical Line-of-Sight (LoS) links between high-altitude UAVs and ground users are frequently obstructed by buildings and other structures. Reconfigurable Intelligent Surfaces (RIS) have emerged as a transformative technology to address this, capable of "bending" wireless signals to create alternative paths and bypass blockages.
Despite their potential, configuring an RIS in real-time is computationally intensive. The system must simultaneously decide which users are allocated to specific RIS elements (a discrete allocation problem) and determine the optimal phase shift for each element (a continuous beamforming problem). This results in a Mixed-Integer Non-Linear Program (MINLP) that is typically too slow to solve for highly dynamic UAV-to-ground links.
Proposed System and Protocol
The authors propose an adaptive RIS-assisted transmission protocol designed for multi-UAV, multi-user scenarios. The system model, illustrated in Figure 1, depicts how an RIS mounted on a building facade can facilitate communication when direct paths are unavailable.
The protocol operates within a frame-based structure divided into two primary phases:
- Negotiation Phase (NP): This phase handles synchronization, channel estimation, and the optimization of RIS configurations. To support multiple users, the RIS is divided into "groups," allowing it to serve multiple UAV-user pairs simultaneously without inter-group interference.
- Communication Phase (CP): Actual data transmission occurs during this phase, utilizing the optimized strategies determined in the NP.
Multi-Task Learning Framework
To overcome the latency of traditional optimization, the researchers developed a Multi-Task Learning (MTL) framework based on a Deep Neural Network (DNN). The core innovation lies in using shared hidden layers that branch into two specialized "heads":
- Classification Head: Predicts the optimal RIS group allocation for users.
- Regression Head: Determines the continuous phase-shift values for the RIS elements.
By training this model offline using data generated from exhaustive searches, the RIS controller can perform near-instantaneous "inference" during the online phase. This approach exploits the underlying physical correlations between allocation and phase-shifting, leading to higher accuracy than training two separate models.
Evaluation and Results
The MTL model was evaluated against traditional Spatial Branch and Bound (sBB) methods using a custom dataset of simulated 3D UAV and user deployments.
- Massive Speedup: The MTL approach reduced computation time by approximately four orders of magnitude (e.g., ~0.017 ms compared to ~3.8 ms for 8 users), enabling near-real-time link enhancement.
- Throughput Gains: In scenarios with blocked LoS, the RIS-assisted system increased throughput from a negligible 0.06 Mbps to 28 Mbps.
- Energy Efficiency: The system significantly reduced the transmit power required by UAVs while maintaining high Signal-to-Noise Ratios (SNR), which is vital for extending the operational life of battery-constrained aerial platforms.
Conclusion
This research demonstrates that integrating RIS with deep learning can effectively mitigate urban signal blockages in aerial-terrestrial networks. By shifting complex optimization to an offline training phase, the proposed MTL framework enables near-real-time configuration of RIS elements. Future work will focus on integrating UAV trajectory planning and addressing the challenges of imperfect channel state information (CSI) in practical, noisy environments.
Summary updated