[Paper Review] Reconfigurable Intelligent Surface-Assisted Aerial-Terrestrial Communications via Multi-Task Learning

Xuelin Cao, Bo Yang+5 more|arXiv (Cornell University)|Apr 14, 2021

Advanced Wireless Communication Technologies57 references4 citations

TL;DR

This paper proposes a reconfigurable intelligent surface (RIS)-assisted transmission protocol for aerial-terrestrial communications using multi-task learning (MTL) to optimize RIS phase shifts and user pairing, significantly improving system throughput and reducing computation time by four orders of magnitude compared to conventional methods.

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.

Motivation & Objective

• Address coverage and link performance degradation in aerial-terrestrial networks due to blockages and path loss in urban environments.
• Overcome the limitations of line-of-sight (LoS) link blockage and high mobility of UAVs in complex propagation environments.
• Optimize RIS configuration and user pairing to maximize system throughput in multi-user downlink scenarios.
• Reduce the computational burden of solving the mixed-integer non-linear programming (MINLP) problem for RIS optimization.

Proposed method

• Design a frame-based RIS-assisted transmission protocol that decouples channel estimation, transmission strategy, and data transmission.
• Formulate the RIS optimization problem as a mixed-integer non-linear program (MINLP) to maximize system throughput.
• Train a multi-task learning (MTL) model at the RIS controller to jointly predict user pairing and RIS phase shift configurations.
• Use MTL to accelerate inference by replacing traditional mathematical optimization, reducing computation time by four orders of magnitude.
• Employ alternating algorithms and spatial branch and bound (sBB) as benchmarks to evaluate MTL performance.
• Leverage end-to-end MTL for joint classification (user pairing) and regression (phase shift optimization) with shared feature representations.

Experimental results

Research questions

• RQ1 How can RIS configuration and user pairing be jointly optimized to maximize throughput in RIS-assisted aerial-terrestrial networks?
• RQ2 To what extent does multi-task learning reduce the computational latency of solving the MINLP optimization problem for RIS deployment?
• RQ3 How does the performance of MTL-based RIS optimization scale with increasing numbers of UAV-user pairs?
• RQ4 What is the trade-off between system throughput and transmit power when using RIS-assisted transmission?
• RQ5 How does the number of training samples affect the inference accuracy and robustness of the MTL model?

Key findings

• The proposed RIS-assisted transmission protocol improves system throughput by up to hundreds of times compared to systems without RIS.
• Multi-task learning reduces the computation time for solving the MINLP problem by approximately four orders of magnitude compared to conventional methods.
• The MTL model maintains high inference accuracy, with classification accuracy declining only slightly as the number of UAV-user pairs increases, outperforming sBB and alternating algorithms.
• The mean square error (MSE) of phase shift regression predictions is significantly lower with MTL than with sBB or alternating algorithms, especially at higher user pair counts.
• Inference time per sample drops from 3.2 ms (sBB) to 0.0145 ms (MTL) when handling two UAV-user pairs, demonstrating massive speedup.
• Performance gains are most pronounced when training data is limited, with MTL showing a 10% to 90% data increase leading to a substantial improvement in classification accuracy, especially for eight UAV-user pairs.