Article Details

Abstract

To address the urgent demand for real-time array resource management in multi-task concurrent scenarios of next-generation integrated TT&C and sensing systems, this paper proposes an adaptive digital array subarray scheduling method based on a physics-aware Graph Neural Network. To tackle the challenge that traditional methods struggle to balance computational efficiency and physical performance when dealing with highly dynamic aerospace targets, the subarray scheduling problem is formulated as a constrained subgraph search problem in graph theory, and a two-stage "prediction + adjustment" algorithmic framework is developed. In the prediction stage, a Graph Neural Network model named GPSM is constructed by incorporating physical prior features. By employing an improved linear attention mechanism and a global feature interaction strategy, large-scale candidate subarrays are efficiently generated. In the local correction stage, a dynamic adjustment strategy is designed to iteratively optimize the candidate node set under beamforming constraints and accuracy constraints, thereby enhancing the feasibility and physical consistency of the scheduling solutions. Simulation experiments are conducted for satellite telemetry, tracking, and sensing tasks. The results demonstrate that the proposed GPSM-CSO algorithm achieves an approximation ratio of 1.13 with respect to the labeled solutions in terms of element utilization, while reducing the average scheduling time by 91.3% and maintaining a scheduling success rate exceeding 99%. Compared with baseline models such as GraphGPS and Transformer, GPSM exhibits significant advantages in prediction accuracy, time efficiency, and F1-score, providing an efficient end-to-end solution for real-time scheduling of digital arrays.

Keywords

Graph Transformer; GNN; adaptive subarray scheduling; digital phased array; TT&C and sensing integration

References

1. [1] WU MANQING. Digital Array Radar: Technology and Trends [C]//Proceedings of 2011 IEEE CIE International Conference on Radar.Chengdu：IEEE，2011：1-4.
2. [2] DELOS P. Advanced Technologies Pave the Way for New Phased Array Radar Architectures [EB/OL].[2025-12-03].https://www.analog.com/en/resources/technical-articles/advanced-technologies-pave-the-way-for-new-phased-array-radar-architectures.html.
3. [3] WAGNER K，GEORGE T. Adaptive Radar Subarray Scheduling [C]//2023 IEEE Radar Conference (RadarConf23).San Antonio：IEEE，2023：1-6.
4. [4] Liu Zijing, Chen Xi, Shi Qingzhan, etc Sub array level adaptive beamforming based on genetic simulated annealing algorithm [J]. Telecommunications Technology, 2024, 64 (9): 1480-1485
5. [5] Li Haoqing, Yu Dian, Pan Changchun, etc Multi radar collaborative task scheduling algorithm based on model knowledge fusion using graph neural network [J]. Journal of Radar (Chinese and English), 2025, 14 (2): 470-485
6. [6] HE Z Y，WANG L，HAO Y. Resource Allocation based on Graph Neural Networks in Vehicular Communications [C]//GLOBECOM 2020-2020 IEEE Global Communications Conference.Taipei：IEEE，2020：1-5.
7. [7] GAO Z，EISEN M，RIBEIRO A，et al. Resource Allocation via Graph Neural Networks in Free Space Optical Fronthaul Networks [C]//GLOBECOM 2020-2020 IEEE Global Communications Conference.Taipei：IEEE，2020：1-6.
8. [8] VANERIO J，CASAS P，SCHMID S. Towards Intelligent Resource Allocation in Highly-Distributed Content Delivery Networks Using Graph Neural Networks [C]//2025 9th Network Traffic Measurement and Analysis Conference (TMA).Copenhagen：IEEE，2025：1-4.
9. [9] SHEN Y F，ZHANG J，SONG S H. Graph Neural Networks for Wireless Communications: From Theory to Practice [J].IEEE Transactions on Wireless Communications，2023，22（5）：3554-3569.
10. [10] CHEN N X，WATANABE S，VILLALBA J，et al. Non-Autoregressive Transformer for Speech Recognition [J].IEEE Signal Processing Letters，2021，28：121-125.
11. [11] DWIVEDI V P，BRESSON X. A Generalization of Transformer Networks to Graphs [J/OL].arXiv:2012.09699 [cs.LG].[2025-12-16].https://doi.org/10.48550/arXiv.2012.09699.
12. [12] YING C X，CAI T L，LUO S J，et al. Do Transformers Really Perform Bad for Graph Representation? [J/OL].arXiv:2106.05234 [cs.LG].[2025-11-17].https://doi.org/10.48550/arXiv.2106.05234.
13. [13] RAMPÁŠEK L，GALKIN M，DWIVEDI V P，et al. Recipe for a General, Powerful, Scalable Graph Transformer [J/OL].arXiv:2205.12454 [cs.LG].[2025-11-16].https://doi.org/10.48550/arXiv.2205.12454.
14. [14] CHOROMANSKI K，VALERII L，DAVID D，et al. Rethinking Attention with Performers [J/OL].arXiv:2009.14794 [cs.LG].[2026-01-02].https://doi.org/10.48550/arXiv.2009.14794.
15. [15] WU H X，WU J L，XU J H，et al. Performer: Linearizing Transformers with Conservation Flows [J/OL].arXiv:2202.06258 [cs.LG].https://doi.org/10.48550/arXiv.2202.06258.
16. [16] LIU Z Y，XU Z H，CHEN S W，et al. Subarray Partition in Array Antenna Based on the Algorithm X [J].IEEE Antennas and Wireless Propagation Letters， 2013，12：906-909.
17. [17] Li Shuopeng, Qi Siyu, Lin Shaofu, etc A review of graph neural networks and their applications in the field of communication networks [J]. Journal of Beijing Institute of Technology, 2021, 47 (8): 971-981
18. [18] BELLO I，ZOPH B，LE Q，et al. Attention Augmented Convolutional Networks [C]//2019 IEEE/CVF International Conference on Computer Vision (ICCV).Seoul：IEEE，2019：3285-3294.
19. [19] YUN S，JEONG M，KIM R，et al. Graph Transformer Networks [C]//Proceedings of the 33rd International Conference on Neural Information Processing Systems.Vancouver：Curran Associates Inc.，2019：11983-11993.

About the authors

Ma Yuqing (2000-), male, Han ethnicity, Beijing native, master's student, research direction: artificial intelligence, graph neural networks; Hu Hefei (1977-), male, Han ethnicity, Chongqing native, associate professor, master's supervisor, research direction: high-speed wireless access networks, software defined networks (SDN).