基于Q学习文化基因算法的考虑任务聚类的多敏捷对地观测卫星调度研究

doi:10.12305/j.issn.1001-506X.2026.04.23

系统工程与电子技术 ›› 2026, Vol. 48 ›› Issue (4): 1349-1359.doi: 10.12305/j.issn.1001-506X.2026.04.23

• 系统工程 • 上一篇

基于Q学习文化基因算法的考虑任务聚类的多敏捷对地观测卫星调度研究

骆婧娴¹, 周光辉¹^,*, 于金月¹, 张扬²

1. 中国科学院大学经济与管理学院，北京 100190
2. 中国科学院空天信息创新研究院，北京 100094

收稿日期:2025-07-02 修回日期:2025-10-02 出版日期:2026-03-20 发布日期:2026-03-20
通讯作者: 周光辉
作者简介:骆婧娴（1998—），女，博士研究生，主要研究方向为航空航天任务规划与管理
于金月（1999—），女，博士研究生，主要研究方向为航空航天任务规划与管理
张　扬（1986—），男，高级工程师，硕士研究生导师，博士，主要研究方向为卫星轨道动力学
基金资助:
国家自然科学基金 (91538113, 72071195, 71402176);中国科学院青年创新促进会（2019171, 2022126）;中国科学院大学数智时代经济管理复杂系统建模教育部哲学社会科学创新团队资助课题

Research on multiple agile earth observation satellite scheduling based on Q-learning memetic algorithm considering task clustering

Jingxian LUO¹, Guanghui ZHOU¹^,*, Jinyue YU¹, Yang ZHANG²

1. School of Economics and Management，University of Chinese Academy of Sciences，Beijing 100190，China
2. Aerospace Information Research Institute，Chinese Academy of Sciences，Beijing 100094，China

Received:2025-07-02 Revised:2025-10-02 Online:2026-03-20 Published:2026-03-20
Contact: Guanghui ZHOU

摘要/Abstract

摘要：

为解决考虑任务聚类的多敏捷对地观测卫星（agile earth observation satellite, AEOS）调度问题，将该问题建模为任务动态聚类、时间依赖、带时间窗的团队定向问题，构建混合整数规划模型，提出一种Q学习文化基因算法（Q-learning memetic algorithm, QLMA），通过基于密度和聚类时间窗的空间点目标聚类算法生成初始种群，设计基于历史聚类信息的交叉算子实现种群的进化。提出一种基于Q学习的邻域选择框架，并设计了AEOS重分配、重聚类、观测任务聚类反转与交换4种邻域搜索算子。不同规模算例的数据实验验证了模型、QLMA以及任务聚类的有效性。

关键词: 卫星调度, 对地观测, 任务聚类, 文化基因算法

Abstract:

To address the multiple agile Earth observation satellite （AEOS） scheduling problem considering task clustering, this problem is modeled as a team oriented problem with task dynamic clustering, time dependent and with time windows. A mixed-integer programming model is constructed. A Q-learning memetic algorithm （QLMA） is proposed, which generates the initial population using a spatial point-target clustering algorithm based on density and clustering time windows. A crossover operator based on historical clustering information is designed to realize population evolution. A Q-learning-based neighborhood selection framework is proposed, with four types of neighborhood search operators—AEOS reallocation, re-clustering, reversal and exchange of observation task clustering. Numerical experiments on instances of different scales validate the effectiveness of the model, the QLMA, and task clustering.

Key words: satellite scheduling, Earth observation, task clustering, memetic algorithm

中图分类号:

O 221

骆婧娴, 周光辉, 于金月, 张扬. 基于Q学习文化基因算法的考虑任务聚类的多敏捷对地观测卫星调度研究[J]. 系统工程与电子技术, 2026, 48(4): 1349-1359.

Jingxian LUO, Guanghui ZHOU, Jinyue YU, Yang ZHANG. Research on multiple agile earth observation satellite scheduling based on Q-learning memetic algorithm considering task clustering[J]. Systems Engineering and Electronics, 2026, 48(4): 1349-1359.

图/表 13

图1

图2

图3

图4

表1

表2

表3

表4

表5

图5

表6

图6

图7

参考文献 29

1	National Aeronautics and Space Administration. Landsat science. [EB/OL]. [2025-04-06]. https: //landsat.gsfc.nasa.gov/.
2	European Space Agency. Europe’s eyes on Earth: the EU’s Copernicus Programme [EB/OL]. [2025-04-08]. https://omcs.atlassian.net/wiki/spaces/CSWK/pages/407404661/Copernicus+Programme.
3	China National Space Administration. High-resolution Earth observation syetem [EB/OL]. [2025-03-09]. https://www.cnsa.gov.cn/n6758824/n6759009/n6759041/c6780463/content.html.
4	LEMAITRE M, VERFAILLIE G, JOUHAUD F, et al. Selecting and scheduling observations of agile satellites[J]. Aerospace Science and Technology, 2002, 6 (5): 367- 381. doi: 10.1016/S1270-9638(02)01173-2
5	WOLFE W J, SORENSEN S E. Three scheduling algorithms applied to the Earth observing systems domain[J]. Management Science, 2000, 46 (1): 148- 166. doi: 10.1287/mnsc.46.1.148.15134
6	WANG Z L, HU X X, MA H W, et al. Learning multi-satellite scheduling policy with heterogeneous graph neural network[J]. Advances in Space Research, 2024, 73 (6): 2921- 2938. doi: 10.1016/j.asr.2023.12.036
7	LI P Y, WANG H Q, ZHANG Y X, et al. Mission planning for distributed multiple agile Earth observing satellites by attention-based deep reinforcement learning method[J]. Advances in Space Research, 2024, 74 (5): 2388- 2404. doi: 10.1016/j.asr.2024.06.003
8	CHANG Z X, ZHOU Z B. Three multi-objective memetic algorithms for observation scheduling problem of active-imaging agile Earth observation satellites[J]. Annals of Operations Research, 2024, 346, 861- 893.
9	CHEN M, DU Y H, TANG K, et al. Learning to construct a solution for the agile satellite scheduling problem with time-dependent transition times[J]. IEEE Trans. on Systems, Man, and Cybernetics: Systems, 2024, 54 (10): 5949- 5963.
10	高天旸, 胡笑旋, 夏维. 基于分布式协同进化的星座自主任务规划算法[J]. 系统工程与电子技术, 2022, 44 (5): 1600- 1608.
	GAO T Y, HU X X, XIA W. Constellation autonomous mission planning algorithm based on distributed co-evolution[J]. Systems Engineering and Electronics, 2022, 44 (5): 1600- 1608.
11	WU G H, LIU J, MA M H, et al. A two-phase scheduling method with the consideration of task clustering for earth observing satellites[J]. Computers & Operations Research, 2013, 40 (7): 1884- 1894.
12	WU G H, WANG H L, WITOLD P, et al. Satellite observation scheduling with a novel adaptive simulated annealing algorithm and a dynamic task clustering strategy[J]. Computers & Industrial Engineering, 2017, 113, 576- 588.
13	WANG P, REINELT G, GAO P, et al. A model, a heuristic and a decision support system to solve the scheduling problem of an Earth observing satellite constellation[J]. Computers & Industrial Engineering, 2011, 61 (2): 322- 335.
14	SPANGELO S, CUTLER J, GILSON K, et al. Optimization-based scheduling for the single-satellite, multi-ground station communication problem[J]. Computers & Operations Research, 2015, 57, 1- 16.
15	PENG G S, DEWIL R, VERBEECK C, et al. Agile Earth observation satellite scheduling: an orienteering problem with time-dependent profits and travel times[J]. Computers & Operations Research, 2019, 111, 84- 98.
16	WOLSEY L A, NEMHAUSER G L. Integer and combinatorial optimization[M]. Hoboken: John Wiley & Sons, 1999.
17	WERTZ J R. Spacecraft attitude determination and control[M]. Berlin: Springer Science & Business Media, 2012.
18	GOLDEN B L, LEVY L, VOHRA R. The orienteering problem[J]. Naval Research Logistics, 1987, 34 (3): 307- 318.
19	LI J J, ZHU J H, PENG G S, et al. Branch-price-and-cut algorithms for the team orienteering problem with interval-varying profits[J]. European Journal of Operational Research, 2024, 319 (3): 793- 807. doi: 10.1016/j.ejor.2024.07.015
20	HE M, WU Q H, BENLIC U, et al. An effective multi-level memetic search with neighborhood reduction for the clustered team orienteering problem[J]. European Journal of Operational Research, 2024, 318 (3): 778- 801. doi: 10.1016/j.ejor.2024.06.015
21	ZHAO J, POON M, TAN V Y F, et al. A hybrid genetic search and dynamic programming-based split algorithm for the multi-trip time-dependent vehicle routing problem[J]. European Journal of Operational Research, 2024, 317 (3): 921- 935. doi: 10.1016/j.ejor.2024.04.011
22	卢弘, 王耀南, 乔非, 等. 面向可持续生产中多任务调度的双重增强模因算法[J]. 自动化学报, 2024, 50 (4): 731- 744.
	LU H, WANG Y N, QIAO F, et al. Dual-enhanced memetic algorithm for multi-task scheduling in sustainable production[J]. Acta Automatica Sinica, 2024, 50 (4): 731- 744.
23	周光辉, 梁伟, 敬帅. 多飞艇协同对地观测和数据传输调度模型与算法研究[J]. 系统工程理论与实践, 2020, 40 (4): 1045- 1056.
	ZHOU G H, LIANG W, JING S. Research on multi-airship cooperative earth observation and data transmission scheduling model and algorithm[J]. Systems Engineering − Theory & Practice, 2020, 40 (4): 1045- 1056.
24	ESTER M, KRIEGEL H P, SANDER J, et al. A density-based algorithm for discovering clusters in large spatial databases with noise[C]//Proc. of the International Conference on Knowledge Discovery and Data Mining, 1996.
25	MICHALEWICZ Z. Genetic algorithms+data structures=evolution programs[M]. Berlin: Springer, 1992.
26	WATKINS C J C H, DAYAN P. Q-learning[J]. Machine Learning, 1992, 8, 279- 292.
27	KARIMI-MAMAGHAN M, MOHAMMADI M, PASDELOUP B, et al. Learning to select operators in meta-heuristics: an integration of Q-learning into the iterated greedy algorithm for the permutation flowshop scheduling problem[J]. European Journal of Operational Research, 2023, 304 (3): 1296- 1330.
28	HU R, HUANG Y F, WU X, et al. Collaborative Q-learning hyper-heuristic evolutionary algorithm for the production and transportation integrated scheduling of silicon electrodes[J]. Swarm and Evolutionary Computation, 2024, 86, 101498. doi: 10.1016/j.swevo.2024.101498
29	LÓPEZ-IBÁÑEZ M, DUBOIS-LACOSTE J, CÁCERES L P, et al. The irace package: iterated racing for automatic algorithm configuration[J]. Operations Research Perspectives, 2016, 3, 43- 58. doi: 10.1016/j.orp.2016.09.002

序号	半长轴/km	轨道倾角/（°）	升交点赤经/（°）	离心率	近地点辐角/（°）
1	7569.35	100.74	83.24	0.0025	10.97
2	7467.87	63.40	338.34	0.0280	4.93
3	7580.23	100.63	39.86	0.0006	260.43
4	7467.80	63.40	116.96	0.0187	4.74
5	7129.37	98.41	280.69	0.0001	60.33
6	7011.36	98.30	84.10	0.0019	218.13
7	7620.68	100.28	190.60	0.0022	30.916
8	7580.05	100.05	283.20	0.0009	207.72

地面站	经纬度
地面站1	（116.98°E, 40.57°N）
地面站2	（75.95°E, 39.51°N）
地面站3	（109.33°E, 18.31°N）
地面站4	（102.80°E, 25.03°N）
地面站5	（20.30°E, 67.90°N）

参数名称	取值范围	调优结果
种群规模	$\{ 10,20,30,40\} $	20
交叉概率	$\{ 0.6,{\text{ }}0.7,{\text{ }}0.8,{\text{ }}0.9\} $	0.6
邻域观测任务临界值	$\{ 2,{\text{ }}3,{\text{ }}4,{\text{ }}5\} $	3
任务分配均衡性阈值	$\{ 0.4,{\text{ }}0.5,{\text{ }}0.6,{\text{ }}0.7\} $	0.5
种群多样性阈值	$\{ 0.5,{\text{ }}0.6,{\text{ }}0.7,{\text{ }}0.8\} $	0.6
最大迭代次数	$\{ 100,200,300,400\} $	300
最大无改进代数	$ \{ 10,20,30,40\} $	20

AEOS数量	观测任务数量	QLMA				Gurobi		QLMA与Gurobi总收益的差值/%
AEOS数量	观测任务数量	最大总收益	平均总收益	最小总收益	计算时间/s	总收益	计算时间/s	QLMA与Gurobi总收益的差值/%
1	10	2.69	2.69	2.69	1.49	2.69	4.63	0.00
	30	7.20	6.97	6.62	5.68	3.03	3600.00	57.92
	50	9.34	8.80	8.47	6.99	−	−	−
	70	11.18	10.48	9.67	9.57	−	−	−
	90	12.25	11.49	10.73	15.72	−	−	−
2	10	4.64	4.64	4.64	2.08	4.64	138.47	0.00
	30	12.58	12.58	12.58	7.31	−	−	−
	50	20.90	19.92	19.00	18.17	−	−	−
	70	26.67	25.78	24.37	26.45	−	−	−
	90	32.73	30.89	29.82	43.77	−	−	−
3	10	4.73	4.73	4.73	2.42	4.73	738.20	0.00
	30	13.28	13.28	13.28	8.81	−	−	−
	50	21.99	21.91	21.86	22.92	−	−	−
	70	32.15	31.75	30.77	58.97	−	−	−
	90	39.13	38.49	37.95	65.66	−	−	−
4	10	5.40	5.40	5.40	3.46	5.40	107.78	0.00
	30	14.84	14.83	14.79	14.39	−	−	−
	50	25.22	25.22	25.21	25.57	−	−	−
	70	36.76	36.74	36.70	39.32	−	−	−
	90	46.90	46.82	46.72	142.84	−	−	−

AEOS数量	观测任务数量	QLMA				MA				QLMA与MA 平均总收益的差值/%
AEOS数量	观测任务数量	最大总收益	平均总收益	最小总收益	计算时间/ s	最大总收益	平均总收益	最小总收益	计算时间/ s	QLMA与MA 平均总收益的差值/%
2	100	35.27	34.03	32.61	56.30	32.36	31.47	30.52	34.46	7.51
	200	46.27	43.56	39.97	64.43	41.52	40.54	39.40	48.95	6.94
	300	52.65	48.68	44.33	146.70	46.46	44.77	43.34	105.96	8.03
	400	55.46	53.50	50.71	203.52	51.02	48.75	46.09	174.86	8.89
	500	57.21	53.70	48.93	201.46	52.39	49.74	44.30	178.31	7.38
4	100	50.57	50.46	50.35	57.53	50.53	50.39	50.10	77.17	0.14
	200	87.02	83.92	81.16	243.35	81.64	80.26	76.55	223.16	4.36
	300	99.91	96.98	91.01	444.22	94.00	91.82	90.12	395.74	5.33
	400	116.39	112.44	108.43	694.62	101.60	99.77	97.96	647.07	11.27
	500	129.05	121.34	116.06	1203.91	113.87	104.98	93.90	609.74	13.48
6	100	50.59	50.57	50.54	65.11	50.57	50.57	50.57	53.65	0.00
	200	99.61	99.40	98.89	325.31	99.26	98.18	97.55	284.81	1.23
	300	135.71	132.15	129.44	722.52	126.55	124.83	122.61	725.40	5.54
	400	157.51	151.76	145.87	1382.75	152.14	148.94	144.17	1422.81	1.86
	500	174.36	171.48	165.94	1938.04	158.76	155.74	152.70	1683.13	9.18
8	100	50.59	50.58	50.57	78.74	50.59	50.58	50.57	60.66	0.00
	200	99.84	99.84	99.83	278.60	99.78	99.42	99.04	250.86	0.42
	300	143.23	141.38	139.98	936.23	138.60	135.34	131.64	794.88	4.27
	400	177.20	173.39	171.39	1620.43	163.37	160.04	158.70	1273.33	7.70
	500	206.38	201.18	197.31	2206.33	185.71	178.91	169.03	1734.53	11.07

基于Q学习文化基因算法的考虑任务聚类的多敏捷对地观测卫星调度研究

Research on multiple agile earth observation satellite scheduling based on Q-learning memetic algorithm considering task clustering

RichHTML

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献 29

相关文章 6

编辑推荐

Metrics

本文评价

AEOS 数量	观测任务数量	QLMA				QLMA_NC				QLMA与QLMA_NC 平均总收益的差值/%
AEOS 数量	观测任务数量	最大总收益	平均总收益	最小总收益	计算时间/ s	最大总收益	平均总收益	最小总收益	计算时间/ s	QLMA与QLMA_NC 平均总收益的差值/%
2	100	35.27	34.03	32.61	56.30	32.77	32.25	31.86	45.03	5.22
	200	46.27	43.56	39.97	64.43	42.65	42.31	41.95	85.45	2.88
	300	52.65	48.68	44.33	146.70	44.59	42.87	40.07	126.72	11.94
	400	55.46	53.50	50.71	203.52	45.66	45.05	44.73	152.96	15.81
	500	57.21	53.70	48.93	201.46	47.39	46.58	45.62	229.01	13.26
4	100	50.57	50.46	50.35	57.53	50.54	50.40	50.45	91.12	0.12
	200	87.02	83.92	81.16	243.35	80.30	78.91	77.15	231.68	5.97
	300	99.91	96.98	91.01	444.22	89.67	87.57	84.09	572.59	9.71
	400	116.39	112.44	108.43	694.62	94.15	91.66	88.98	624.79	18.48
	500	129.05	121.34	116.06	1203.91	95.06	93.79	91.55	643.49	22.71
6	100	50.59	50.58	50.54	65.11	50.59	50.57	50.54	80.86	0.01
	200	99.61	99.40	98.89	325.31	98.81	97.23	96.00	350.40	2.18
	300	135.71	132.15	129.44	722.52	118.78	117.53	116.17	666.29	11.06
	400	157.51	151.76	145.87	1382.75	139.96	132.91	125.64	1157.04	12.42
	500	174.36	171.48	165.94	1938.04	141.67	139.16	136.41	1907.67	18.85
8	100	50.59	50.58	50.57	78.74	50.59	50.58	50.57	83.85	0.00
	200	99.84	99.84	99.83	278.60	99.84	99.77	99.64	284.40	0.06
	300	143.23	141.38	139.98	936.23	136.20	134.43	131.60	968.10	4.92
	400	177.20	173.39	171.39	1620.43	156.97	154.16	150.19	1427.18	11.09
	500	206.38	201.18	197.31	2206.33	172.74	167.35	161.33	1946.55	16.82

[1]	靳鹏, 唐晓茜. 考虑合成机制的多星应急任务调度[J]. 系统工程与电子技术, 2022, 44(4): 1270-1281.
[2]	李德伟, 江利明, 蒋厚军, 董景龙, 汪汉胜. 月基SAR对地观测系统参数分析[J]. 系统工程与电子技术, 2020, 42(4): 792-798.
[3]	姜维，郝会成，李一军. 对地观测卫星任务规划问题研究述评[J]. 系统工程与电子技术, 2013, 35(9): 1878-1885.
[4]	邱涤珊，王建江，吴朝波，刘刚. 基于任务合成的对地观测卫星应急调度方法[J]. Journal of Systems Engineering and Electronics, 2013, 35(7): 1430-1437.
[5]	江建清，朱晓敏，伍国华，贺川，吴朝波. 面向应急对地观测任务的多飞艇协同分配方法[J]. Journal of Systems Engineering and Electronics, 2013, 35(2): 317-325.
[6]	郭浩，伍国华，邱涤珊. 敏捷成像卫星密集任务聚类方法[J]. Journal of Systems Engineering and Electronics, 2012, 34(5): 931-935.