Research Article
Secure and Energy-Efficient Low-Altitude Wireless Networks via Joint UAV and STAR-RIS Beamforming: A Heterogeneous Multi-agent DRL Approach
@INPROCEEDINGS{10.1007/978-3-032-44094-5_8, author={Xinyue Liang and Hui Kang and Jiahui Li and Geng Sun and Jiacheng Wang and Changyuan Zhao and Dusit Niyato}, title={Secure and Energy-Efficient Low-Altitude Wireless Networks via Joint UAV and STAR-RIS Beamforming: A Heterogeneous Multi-agent DRL Approach}, proceedings={Industrial Networks and Intelligent Systems. 12th EAI International Conference, INISCOM 2026, Da Nang, Vietnam, February 26--27, 2026, Proceedings}, proceedings_a={INISCOM}, year={2026}, month={11}, keywords={UAV STAR-RIS Secure communications Multi-agent deep reinforcement learning Collaborative beamforming}, doi={10.1007/978-3-032-44094-5_8} }- Xinyue Liang
Hui Kang
Jiahui Li
Geng Sun
Jiacheng Wang
Changyuan Zhao
Dusit Niyato
Year: 2026
Secure and Energy-Efficient Low-Altitude Wireless Networks via Joint UAV and STAR-RIS Beamforming: A Heterogeneous Multi-agent DRL Approach
INISCOM
Springer
DOI: 10.1007/978-3-032-44094-5_8
Abstract
Low-altitude wireless networks (LAWNs) based on uncrewed aerial vehicles (UAVs) show promising capabilities for urban communications due to their mobility, flexible deployment, and coverage advantages. However, dense urban environments cause severe signal attenuation from obstructed propagation paths. To address this issue, we propose integrating collaborative beamforming (CB) of UAVs with omnidirectional reconfigurable beamforming (ORB) of simultaneous transmitting and reflecting reconfigurable intelligent surfaces (STAR-RIS). Specifically, we consider a STAR-RIS-assisted UAV-enabled virtual antenna array (UVAA) system with jointly optimized CB and ORB, and formulate a rate-energy joint optimization problem (REJOP) that maximizes transmission rate while minimizing energy consumption. To solve this non-convex, NP-hard problem, we develop a heterogeneous multi-agent collaborative dynamic (HMCD) optimization framework with two core components, namely a simulated annealing (SA)-based STAR-RIS control method for dynamic coefficient optimization, and an improved multi-agent deep reinforcement learning (MADRL) control method featuring self-attention evaluation and adaptive velocity transition mechanisms. Simulations demonstrate superior performance of HMCD over baselines in convergence speed, transmission rate, and energy efficiency, with transmission rates scaling positively with UAV count.