Global Navigation Satellite System (GNSS) environment recognition is important for enhancing positioning reliability and context awareness in complex urban and natural scenes. However, existing methods predominantly rely on signal features and necessitate extensive labeled datasets, compromising their robustness in label-scarce scenarios. To address this, we propose a large language model (LLM)-assisted, few-sample GNSS environment recognition framework. Our approach leverages LLM-generated physical-semantic soft labels for knowledge distillation during training and incorporates a lightweight test-time adaptation (TTA) strategy during inference. Experimental results demonstrate that with only 1% of labeled training data, the proposed method outperforms the baseline by 6% in classification accuracy.
Qi Yao, Tong Yang, Jingjing Yang et al.· IEEE Wireless Communications...· 0 citations
Reconfigurable intelligent surfaces (RISs) and six-dimensional movable antennas (6DMAs) have emerged as promising technologies for enhancing wireless transmission by reconfiguring the propagation environment and antenna geometry, respectively. However, conventional RIS-assisted systems usually rely on fixed-position antenna arrays, whose spatial adaptability is limited when users are non-uniformly distributed in three-dimensional space. To address this limitation, this paper investigates a STAR-RIS-assisted six-dimensional movable antenna base station (6DMA-BS) communication system. In the considered system, the 6DMA-BS adjusts the positions and rotations of multiple antenna surfaces to exploit additional spatial degrees of freedom, while the STAR-RIS provides full-space coverage for users in both reflection and transmission zones. To enhance the system sum rate, a joint optimization problem is formulated by considering 6DMA surface positions and rotations, BS active beamforming, and STAR-RIS passive reflection/transmission beamforming under practical constraints. Since the formulated problem is highly non-convex with strongly coupled variables, an alternating optimization framework is developed. Specifically, the position and rotation subproblems are solved by conditional gradient methods, the active beamforming is obtained by MRT-based direction design and water-filling-based power allocation, and the passive beamforming is optimized via SDR and Gaussian randomization. Simulation results show that the proposed method can significantly improve the sum rate of users as compared to benchmark with FPA + Optimized STAR-RIS, 6DMA + Random STAR-RIS, and FPA + Random STAR-RIS.
Yuewei Wu, Minghao Chen, Jingjing Yang et al.· IEEE Open Journal of the Com...· 0 citations