基于混合Jaya-Transformer的超敏捷卫星任务调度算法

doi:10.16182/j.issn1004731x.joss.26-0395

摘要/Abstract

摘要：

超敏捷卫星任务调度具有约束复杂、任务转换关系紧密等特点，为解决传统启发式算法在线求解时间较长的问题，建立了以任务收益率最大化为目标，综合考虑可见时间窗口、姿态转换时间、能量和存储容量约束的超敏捷卫星任务调度模型；利用Transformer学习任务需求、可见时间窗口及任务间全局关联关系，结合动态掩码和任务调度器生成可行任务序列。为提高模型对复杂任务关联关系的学习能力，降低人工调参对算法性能的影响，提出一种混合Jaya-Transformer调度算法。实验结果表明：混合Jaya-Transformer相较于Transformer的平均任务收益率提高2.58~11.89个百分点；与ALNS和IGA-LS相比，在保持最优或近似最优任务收益率的同时平均求解时间仅为0.21~0.78 s，低于启发式算法的6.21~108.9 s，能够兼顾调度质量与求解效率。

关键词: 超敏捷卫星, Transformer, 超参数优化, 混合Jaya算法, 任务调度

Abstract:

Super-agile satellite task scheduling is characterized by complex constraints and tight coupling among task transitions. To address the relatively long online solution times of traditional heuristic algorithms, this paper establishes a super-agile satellite task scheduling model that aims to maximize the task profit ratio while considering constraints on visible time windows, attitude transition time, energy, and storage capacity. A Transformer is used to learn task requirements, visible time windows, and global correlations among tasks, and a dynamic mask and a task scheduler are incorporated to generate feasible task sequences. To enhance the model's ability to learn complex task correlations and reduce the influence of manual hyperparameter tuning on algorithm performance, this paper proposes a hybrid Jaya-Transformer scheduling algorithm. Experimental results show that the hybrid Jaya-Transformer improves the average task profit ratio by 2.58-11.89 percentage points compared with Transformer. Compared with ALNS and IGA-LS, the proposed algorithm achieves optimal or near-optimal task profit ratios with an average solution time of only 0.21-0.78 s, substantially shorter than the 6.21-108.9 s required by the heuristic algorithms, demonstrating its ability to balance scheduling quality and solution efficiency.

Key words: super-agile satellite, Transformer, hyperparameter optimization, hybrid Jaya algorithm, task scheduling

中图分类号:

TP391.9

赵浩宇,张广辉,郝兰婷等 . 基于混合Jaya-Transformer的超敏捷卫星任务调度算法[J]. 系统仿真学报, 2026, 38(10): 3052-3067.

Zhao Haoyu,Zhang Guanghui,Hao Lanting,et al . Hybrid Jaya-Transformer-based Task Scheduling Algorithm for Super-agile Satellites[J]. Journal of System Simulation, 2026, 38(10): 3052-3067.

图/表 8

图1

图2

表1

图3

图4

表2

表3

图5

参考文献 24

[1]	Ferrari Benedetta, Cordeau Jean-François, Delorme Maxence, et al. Satellite Scheduling Problems: A Survey of Applications in Earth and Outer Space Observation[J]. Computers & Operations Research, 2025, 173: 106875.
[2]	Wu Guohua, Luo Qizhang, Du Xiao, et al. Ensemble of Metaheuristic and Exact Algorithm Based on the Divide-and-conquer Framework for Multisatellite Observation Scheduling[J]. IEEE Transactions on Aerospace and Electronic Systems, 2022, 58(5): 4396-4408.
[3]	Wang Jianjiang, Zhu Xiaomin, Qiu Dishan, et al. Dynamic Scheduling for Emergency Tasks on Distributed Imaging Satellites with Task Merging[J]. IEEE Transactions on Parallel and Distributed Systems, 2014, 25(9): 2275-2285.
[4]	Wei Luona, Xing Lining, Wan Qian, et al. A Multi-objective Memetic Approach for Time-dependent Agile Earth Observation Satellite Scheduling Problem[J]. Computers & Industrial Engineering, 2021, 159: 107530.
[5]	毛成辉, 胡淦, 顾轶, 等. 基于改进自适应大邻域搜索的低轨预警卫星调度算法[J/OL]. 控制理论与应用. (2025-07-17)[2026-03-15]. .
	Mao Chenghui, Hu Gan, Gu Yi, et al. Low Earth Orbit Early Warning Satellite Scheduling Algorithm Based on Improved Adaptive Large Neighborhood Search[J/OL]. Control Theory & Applications. (2025-07-17) [2026-03-15]. .
[6]	李庥甜, 王凌, 陈英武, 等. 基于自适应大邻域搜索的多场景多卫星任务规划方法[J]. 系统仿真学报, 2025, 37(7): 1836-1847.
	Li Xiutian, Wang Ling, Chen Yingwu, et al. Multi-scenario Multi-satellite Mission Planning Method Based on Adaptive Large Neighborhood Search[J]. Journal of System Simulation, 2025, 37(7): 1836-1847.
[7]	Sun Xiaohan, Ren Yuan, Yu Linghui. Adaptive-strategies-based Quantum Genetic Algorithm for Agile Earth Observation Satellite Scheduling Problem[J]. IEEE Aerospace and Electronic Systems Magazine, 2025, 40(7): 4-15.
[8]	张广辉, 魏晨轩, 冯彦翔, 等. 基于分布式元强化学习的多敏捷卫星多目标调度算法[J]. 控制与决策, 2026, 41(4): 1065-1076.
	Zhang Guanghui, Wei Chenxuan, Feng Yanxiang, et al. Multiple Agile Satellites Multi-objective Scheduling Algorithm Based on Distributed Meta-reinforcement Learning[J]. Control and Decision, 2026, 41(4): 1065-1076.
[9]	张广辉, 魏晨轩, 冯彦翔, 等. 基于混合进化博弈的多敏捷卫星多目标调度优化[J]. 控制理论与应用, 2025, 42(11): 2341-2351.
	Zhang Guanghui, Wei Chenxuan, Feng Yanxiang, et al. Multi-agile Satellites Multi-objective Scheduling Based on Hybrid Evolutionary Game Theory[J]. Control Theory & Applications, 2025, 42(11): 2341-2351.
[10]	Lu Zezhong, Shen Xin, Li Deren, et al. A Mission Planning Modeling Method of Multipoint Target Imaging Within a Single Pass for Super-agile Earth Observation Satellite[J]. IEEE Systems Journal, 2022, 16(2): 1921-1932.
[11]	Lu Zezhong, Shen Xin, Li Deren, et al. Integrated Imaging Mission Planning Modeling Method for Multi-type Targets for Super-agile Earth Observation Satellite[J]. IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, 2022, 15: 4156-4169.
[12]	Wu Guohua, Xiang Zhiqing, Wang Yalin, et al. Improved Adaptive Large Neighborhood Search Algorithm Based on the Two-stage Framework for Scheduling Multiple Super-agile Satellites[J]. IEEE Transactions on Aerospace and Electronic Systems, 2024, 60(5): 7185-7200.
[13]	Caleiras Margarida, Moniz Samuel, Paulo Jorge Nascimento. A Constraint Programming Model for the Super-agile Earth Observation Satellite Imaging Scheduling Problem[EB/OL]. (2026-01-21) [2026-02-09]. .
[14]	Bresson Xavier, Laurent Thomas. The Transformer network for the traveling salesman problem[EB/OL]. (2021-03-04)[2025-12-09]. .
[15]	Liu Zheng, Xiong Wei, Han Chi, et al. A Multi-pointer Network for Multiple Agile Optical Satellite Scheduling Problem[J]. Aerospace, 2024, 11(10): 792.
[16]	Liu Zheng, Wei Xiong, Jian Ping. An Agile Optical Satellite Scheduling Method Based on Transformer Networks[C]//Third International Conference on Electronic Information Engineering, Big Data, and Computer Technology (EIBDCT 2024). Bellingham: SPIE, 2024: 131811E.
[17]	Goismi Mohamed, Debbab Mohamed, Maaskri Moustafa, et al. Enhanced Spam Detection Using Mutual Information Feature Selection and JAYA-tuned Tab Transformer[J]. Journal of Engineering and Technology for Industrial Applications, 2026, 12(57): 1084-1093.
[18]	Dash Nitu, Chakravarty Sujata, Amiya Kumar Rath, et al. An Optimized LSTM-based Deep Learning Model for Anomaly Network Intrusion Detection[J]. Scientific Reports, 2025, 15(1): 1554.
[19]	Gundu Venkateswarlu, Simon Sishaj P, Kumba Krishna. Short-term Solar Power Forecasting—An Approach Using JAYA Based Recurrent Network Model[J]. Multimedia Tools and Applications, 2024, 83(11): 32411-32422.
[20]	Vaswani A, Shazeer N, Parmar N, et al. Attention is All You Need[C]//Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook: Curran Associates Inc., 2017: 6000-6010.
[21]	Xu Shengwei, Han Jieie, Liu Yilong, et al. Few-shot Traffic Classification Based on Autoencoder and Deep Graph Convolutional Networks[J]. Scientific Reports, 2025, 15(1): 8995.
[22]	Adak Sumit, Witt Carsten. Mathematical Runtime Analysis of a Multi-valued Estimation of Distribution Algorithm[J]. Artificial Intelligence, 2026, 353: 104501.
[23]	Liu Xiaolu, Laporte Gilbert, Chen Yingwu, et al. An Adaptive Large Neighborhood Search Metaheuristic for Agile Satellite Scheduling with Time-dependent Transition Time[J]. Computers & Operations Research, 2017, 86: 41-53.
[24]	Song Yanjie, Zhang Zhongshan, Song Bingyu, et al. Improved Genetic Algorithm with Local Search for Satellite Range Scheduling System and Its Application in Environmental monitoring[J]. Sustainable Computing: Informatics and Systems, 2019, 21: 19-27.

卫星参数	值
半长轴/km	6 678.14
偏心率	0
轨道倾角/(°)	97.406 5
升交点赤经/(°)	120
近地点幅角/(°)	0
真近点角/(°)	34
时间	2025-04-20T04:00:00—21T04:00:00

任务数目	算法	任务收益率/%			平均求解时间/s
任务数目	算法	最小值	最大值	平均值	平均求解时间/s
50	ALNS	99.21	100	99.69	8.34
	IGA-LS	96.85	100	99.22	6.21
	Transformer	96.55	97.68	97.07	0.20
	本文	98.26	100	99.65	0.21
100	ALNS	90.58	94.38	92.32	29.74
	IGA-LS	85.58	94.57	90.34	24.34
	Transformer	87.23	89.94	88.03	0.40
	本文	89.10	94.94	93.24	0.45
150	ALNS	80.25	84.65	82.33	67.43
	IGA-LS	75.35	83.40	79.52	67.80
	Transformer	69.24	72.36	71.27	0.59
	本文	81.73	85.68	83.16	0.61
200	ALNS	70.98	75.23	72.48	106.42
	IGA-LS	63.31	74.68	70.71	108.90
	Transformer	63.92	65.22	64.41	0.84
	本文	70.33	77.41	74.36	0.78

任务数目	算法	标准差/%	p-value	显著性
50	ALNS	0.25	0.556 4	ns
	IGA-LS	1.33	0.345 2	ns
	Transformer	0.37	0.002 9	**
	本文	0.62	--	--
100	ALNS	1.27	0.042	*
	IGA-LS	2.67	0.037 1	*
	Transformer	0.80	0.002 9	**
	本文	1.66	--	--
150	ALNS	1.63	0.065 4	ns
	IGA-LS	2.32	0.013 7	*
	Transformer	1.00	0.002 9	**
	本文	1.21	--	--
200	ALNS	1.34	0.037 1	*
	IGA-LS	3.60	0.037 1	*
	Transformer	0.44	0.002 9	**
	本文	2.07	--	--