非完备感知条件下的地面多目标行为与意图预测

朱心怡; 平鹏; 侯婉莹; 施佺; 吴奇

doi:10.11999/JEIT250322

非完备感知条件下的地面多目标行为与意图预测

doi: 10.11999/JEIT250322 cstr: 32379.14.JEIT250322

朱心怡¹,
平鹏^1, ,,
侯婉莹²,
施佺¹,
吴奇³

1.
南通大学交通与土木工程学院南通 226001
2.
中国科学院空天信息创新研究院北京 100094
3.
上海交通大学计算机科学与工程系上海 200240

基金项目: 国家自然科学基金(52442218, U2433216, 52202496)

详细信息

作者简介:
朱心怡：女，硕士生，研究方向为人工智能、态势感知等

平鹏：男，副教授，研究方向为自动驾驶感知技术、人工智能理论研究等

侯婉莹：女，工程师，研究方向为世界模型、视觉推理等

施佺：男，教授，研究方向为智能信息处理、交通大数据分析、无人驾驶、交通信息及控制等

吴奇：男，教授，研究方向为脑科学、类脑计算与人因工程、智慧座舱人机交互、机器人、人机对抗等

通讯作者:
平鹏　pingpeng@ntu.edu.cn

中图分类号: TN929.5; TP391
计量
- 文章访问数: 16
- HTML全文浏览量: 6
- PDF下载量: 1
- 被引次数: 0
出版历程
- 收稿日期: 2025-04-27
- 修回日期: 2025-08-28
- 网络出版日期: 2025-09-02

Multi-target Behavior and Intent Prediction on the Ground Under Incomplete Perception Conditions

ZHU Xinyi¹,
PING Peng^{1
, ,},
HOU Wanying²,
SHI Quan¹,
WU Qi³

1.
School of Transportation and Civil Engineering, Nantong University, Nantong, Jiangsu 226001, China
2.
Aerospace Information Research Institute, Chinese Academy of Sciences, Beijing 100094, China
3.
Department of Computer Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China

Funds: The National Natural Science Foundation of China (52442218, U2433216, 52202496)

摘要

摘要: 现代战场中，目标行为的复杂构成与演化不确定性显著增加了意图预测的难度。传统意图预测方法数据缺失的鲁棒应对不足、目标行为模态考虑较为固化，易受复杂战场环境影响，难以适应快速变化战场环境下的高价值目标意图识别与整体地面目标的态势感知。为此，该文提出一种融合威胁场建模与动态修复机制的门控循环单元(GRU)预测模型(TF-GRU)。该模型首先构建静态威胁场与动态威胁场以关联目标特性与意图，继而通过粒子滤波与动态时间规整的动态融合策略处理数据缺失，并引入邻域目标角度约束增强多目标预测能力，继而将轨迹数据与威胁场数据输入GRU捕捉目标行为的时序动态演化，从而在信息非完备条件下实现对地面目标整体意图的精准预测。实验结果表明，该方法显著提高了意图预测的准确性，可为战场态势感知和决策提供强有力的支持。
- 意图预测 /
- 态势感知 /
- 粒子滤波 /
- 门控循环单元 /
- 威胁场
Abstract: Objective Modern battlefield environments, characterized by complex and dynamically uncertain target behaviors combined with information asymmetry, present significant challenges for intent prediction. Conventional methods lack robustness in processing incomplete data, rely on oversimplified behavioral models, and fail to capture tactical intent semantics or adapt to rapidly evolving multi-target coordinated scenarios. These limitations restrict their ability to meet the demands of real-time recognition of high-value target intent and comprehensive ground target situational awareness. To address these challenges, this study proposes a Threat Field-integrated Gated Recurrent Unit model (TF-GRU), which improves prediction accuracy and robustness through threat field modeling, dynamic data repair, and multi-target collaboration, thereby providing reliable support for battlefield decision-making. Methods The TF-GRU framework integrates static and dynamic threat field modeling with a hybrid Particle Filtering (PF) and Dynamic Time Warping (DTW) strategy. Static threat fields quantify target-specific threats (e.g., tanks, armored vehicles, artillery) using five factors: enemy–friend distance, range, firepower, defense, and mobility. Gaussian and exponential decay models are employed to describe spatial threat diffusion across different target categories. Dynamic threat fields incorporate real-time kinematic variables (velocity, acceleration, orientation) and temporal decay, allowing adaptive updates of threat intensity. To address incomplete sensor data, a PF–DTW switching mechanism dynamically alternates between short-term PF (N = 1,000 particles) and long-term historical trajectory matching (DTW with β = 50). Collaborative PF introduces neighborhood angular constraints to refine multi-target state estimation. The GRU architecture is further enhanced with Mish activation, adaptive Xavier initialization, and threat-adaptive gating, ensuring effective fusion of trajectory and threat features. Results and Discussions Experiments were conducted on a simulated dataset comprising 150 trajectories and 270,000 timesteps. Under complete data conditions, the TF-GRU model achieved the highest accuracy on both the training and test sets, reaching 94.7% and 92.9%, respectively, indicating strong fitting capability and generalization performance (Fig. 10). After integrating static and dynamic threat fields, model accuracy increased from 72% (trajectory-only input) to 83%, accompanied by substantial improvements in F1 scores and reductions in predictive uncertainty (Fig. 7). In scenarios with missing data, TF-GRU maintained an accuracy of 86.2%, outperforming comparative models and demonstrating superior robustness (Fig. 11). These results confirm that the PF–DTW mechanism effectively reduces the adverse effects of both short-term and long-term data loss, while the collaborative PF strategy strengthens multi-target prediction through neighborhood synergy (η = 0.6). This combination enables robust threat field reconstruction and reliable intent inference (Figs. 8–9). Conclusions The TF-GRU model effectively addresses the challenges of intent prediction in complex battlefield environments with incomplete data through threat field modeling, the PF–DTW dynamic repair mechanism, and multi-target collaboration. It achieves high accuracy and robustness, providing reliable support for situational awareness and command decision-making. Future work will focus on applying the model to real-world datasets and enhancing computational efficiency to facilitate practical deployment.
- Intent prediction /
- Situational awareness /
- Particle Filtering (PF) /
- Gated Recurrent Unit (GRU) /
- Threat field

HTML全文

图 1 3种车辆类型的威胁强度变化

下载: 全尺寸图片幻灯片

图 2 威胁场关键要素图

下载: 全尺寸图片幻灯片

图 3 威胁场建模示意图

下载: 全尺寸图片幻灯片

图 4 多目标威胁场演变示意图

下载: 全尺寸图片幻灯片

图 5 基于目标轨迹和威胁场的单目标GRU模型框架

下载: 全尺寸图片幻灯片

图 6 静态与动态威胁场的有效性验证

下载: 全尺寸图片幻灯片

图 7 粒子数与协同系数的选定

下载: 全尺寸图片幻灯片

图 8 不同缺失比例下模型性能对比

下载: 全尺寸图片幻灯片

图 9 TF-GRU模型实验结果

下载: 全尺寸图片幻灯片

图 10 不同缺失率下的准确率对比

下载: 全尺寸图片幻灯片

表 1 动态威胁因子定义

威胁因子	描述	归一化函数
$ v $	速度	$ {C}_{v}=\frac{1}{1+{\mathrm{e}}^{-\left(\frac{{v}_{\left\|\right\|}\left(t\right)-{v}_{0}}{{b}_{v}}\right)}} $
$ a $	加速度	$ {C}_{a}=\frac{1}{1+{\mathrm{e}}^{-\left(\frac{{a}_{\left\|\right\|}\left(t\right)-{a}_{0}}{{b}_{a}}\right)}} $
$ \theta $	朝向	$ {C}_{\theta }=1-\dfrac{\left\|\theta \right(t\left)\right\|}{{180}^{\circ }} $
$ R $	距离	$ {C}_{R}={\mathrm{e}}^{-{\left(\frac{R\left(t\right)-{R}_{0}}{{b}_{R}}\right)}^{2}} $
注：$ {v}_{0} $是基准速度，通常为0，$ {b}_{v} $控制速度对威胁度的敏感程度; $ {a}_{0} $为基准加速度值，$ {b}_{a} $反映加速度的变化速度对威胁度的影响程度；$ {R}_{0} $为参考距离，是目标的最大射程或目标威胁显著增强的距离，$ {b}_{R} $控制距离对威胁的衰减速度。

下载: 导出CSV

表 2 目标特性与威胁参数设定

目标类型	数量 (辆)	射程 (m)	火力	防御	机动	最优射程(m)	重要性$ {\omega }_{i} $	威胁扩散系数$ {\sigma }_{i} $
坦克	2	3 000	0.9	0.8	0.6	1 500	0.5	300
装甲车	2	1 000	0.6	0.5	0.8	500	0.2	100
火炮	1	8 000	0.7	0.4	0.4	5 000	0.3	800

下载: 导出CSV

表 3 意图编码

意图类型	编码
进攻	0
侦察	1
撤收	2
行动	3
打击	4

下载: 导出CSV

表 4 模型参数

参数	TF-GRU	CNN-LSTM	Transformer
输入维度	6	6	6
特征增强	2	—	—
空间特征提取	威胁场建模	2层1D-CNN(通道32→64)	多头自注意力(4头)
时序建模	改进型GRU(隐藏层128)	LSTM(隐藏层64)	Transformer编码器(4层，隐藏层128)
优化器	Adam (lr=0.001)	Adam (lr=0.001)	Adam (lr=0.001)
训练周期	200	200	200
损失函数	加权交叉熵	加权交叉熵	加权交叉熵

下载: 导出CSV

参考文献(33)

[1]	ENDSLEY M R. Toward a theory of situation awareness in dynamic systems[J]. Human Factors, 1995, 37(1): 32–64. doi: 10.1518/001872095779049543.
[2]	ENDSLEY M R. Design and evaluation for situation awareness enhancement[J]. Proceedings of the Human Factors Society Annual Meeting, 1988, 32(2): 97–101. doi: 10.1177/154193128803200221.
[3]	KOTT A and MCENEANEY W M. Adversarial Reasoning: Computational Approaches to Reading the Opponent's Mind[M]. New York: Chapman and Hall/CRC, 2006: 340. doi: 10.1201/9781420011012.
[4]	GMYTRASIEWICZ P J and DOSHI P. A framework for sequential planning in multi-agent settings[J]. Journal of Artificial Intelligence Research, 2005, 24: 49–79. doi: 10.1613/jair.1579.
[5]	GOODRICH M A and SCHULTZ A C. Human-robot interaction: A survey[J]. Foundations and Trends in Human-Computer Interaction, 2007, 1(3): 203–275. doi: 10.1561/1100000005.
[6]	王坤, 丁麒龙. 利用自适应融合和混合锚检测器的遥感图像小目标检测算法[J]. 电子与信息学报, 2024, 46(7): 2942–2951. doi: 10.11999/JEIT230966. WANG Kun and DING Qilong. Remote sensing images small object detection algorithm with adaptive fusion and hybrid anchor detector[J]. Journal of Electronics & Information Technology, 2024, 46(7): 2942–2951. doi: 10.11999/JEIT230966.
[7]	肖振久, 张杰浩, 林渤翰. 特征协同与细粒度感知的遥感图像小目标检测[J]. 光电工程, 2024, 51(6): 240066. doi: 10.12086/oee.2024.240066. XIAO Zhenjiu, ZHANG Jiehao, and LIN Bohan. Feature coordination and fine-grained perception of small targets in remote sensing images[J]. Opto-Electronic Engineering, 2024, 51(6): 240066. doi: 10.12086/oee.2024.240066.
[8]	贾亮, 林铭文, 戚丽瑾, 等. 面向无人机航拍图像的多尺度目标检测研究[J]. 半导体光电, 2024, 45(3): 501–507,514. doi: 10.16818/j.issn1001-5868.2024010903. JIA Liang, LIN Mingwen, QI Lijin, et al. Multiscale target detection for UAV aerial images[J]. Semiconductor Optoelectronics, 2024, 45(3): 501–507,514. doi: 10.16818/j.issn1001-5868.2024010903.
[9]	黄玉玲, 陶昕辰, 朱涛, 等. 残差对抗目标检测算法的遥感图像检测[J]. 电光与控制, 2023, 30(7): 63–67. doi: 10.3969/j.issn.1671-637X.2023.07.011. HUANG Yuling, TAO Xinchen, ZHU Tao, et al. A remote sensing image detection method based on residuals adversarial object detection algorithm[J]. Electronics Optics & Control, 2023, 30(7): 63–67. doi: 10.3969/j.issn.1671-637X.2023.07.011.
[10]	段玉先, 刘昌云, 魏文凤. 战场态势感知关键技术研究[J]. 火力与指挥控制, 2021, 46(11): 1–11,19. doi: 10.3969/j.issn.1002-0640.2021.11.001. DUAN Yuxian, LIU Changyun, and WEI Wenfeng. Review of key technologies for battlefield situational awareness[J]. Fire Control & Command Control, 2021, 46(11): 1–11,19. doi: 10.3969/j.issn.1002-0640.2021.11.001.
[11]	LI Lemin, SONG Yafei, WANG Ke, et al. BiLSTM-TCN: An aerial target intent recognition model[C]. 2023 China Automation Congress (CAC), Chongqing, China, 2023: 1164–1170. doi: 10.1109/CAC59555.2023.10451474.
[12]	龚诚, 衣军波, 蔡林恭, 等. 基于知识图谱嵌入学习的目标意图智能识别方法[J]. 舰船电子工程, 2023, 43(1): 32–35,40. doi: 10.3969/j.issn.1672-9730.2023.01.008. GONG Cheng, YI Junbo, CAI Lingong, et al. Intelligent recognition method of target intention based on knowledge graph embedded learning[J]. Ship Electronic Engineering, 2023, 43(1): 32–35,40. doi: 10.3969/j.issn.1672-9730.2023.01.008.
[13]	陈优敏, 李长云. 基于知识图谱的目标战术意图识别仿真[J]. 计算机仿真, 2019, 36(8): 1–4,19. doi: 10.3969/j.issn.1006-9348.2019.08.001. CHEN Youmin and LI Changyun. Simulation of target tactical intention recognition based on knowledge map[J]. Computer Simulation, 2019, 36(8): 1–4,19. doi: 10.3969/j.issn.1006-9348.2019.08.001.
[14]	尹翔, 张萌, 陈梦乔. 基于判别分析的空中目标作战意图识别[J]. 弹箭与制导学报, 2018, 38(3): 46–50. doi: 10.15892/j.cnki.djzdxb.2018.03.011. YIN Xiang, ZHANG Meng, and CHEN Mengqiao. Combat intention recognition of the target in the air based on discriminant analysis[J]. Journal of Projectiles, Rockets, Missiles and Guidance, 2018, 38(3): 46–50. doi: 10.15892/j.cnki.djzdxb.2018.03.011.
[15]	KIM P S, HYUN S Y, and HA Y G. Battlefield situation awareness model using convolutional LSTM[C]. 2023 IEEE International Conference on Big Data and Smart Computing (BigComp), Jeju, Korea, Republic of, 2023: 369–370. doi: 10.1109/BigComp57234.2023.00086.
[16]	KIM A and LEE D. Dynamic Bayesian network-based situational awareness and course of action decision-making support model[J]. Expert Systems with Applications, 2024, 252: 124093. doi: 10.1016/j.eswa.2024.124093.
[17]	王宇航, 董宝良, 公超, 等. 基于意图识别的空中群目标动态威胁评估[J]. 计算机与现代化, 2023(12): 100–104, 111. doi: 10.3969/j.issn.1006-2475.2023.12.017. WANG Yuhang, DONG Baoliang, GONG Chao, et al. Dynamic threat assessment of air swarm targets based on intent recognition[J]. Computer and Modernization, 2023(12): 100–104+111. doi: 10.3969/j.issn.1006-2475.2023.12.017.
[18]	FOERSTER J N, ASSAEL Y M, DE FREITAS N, et al. Learning to communicate with deep multi-agent reinforcement learning[C]. Proceedings of the 30th International Conference on Neural Information Processing Systems, Barcelona, Spain, 2016: 2145–2153.
[19]	LIU Yongkang, QI Xuewei, SISBOT E A, et al. Multi-agent trajectory prediction with graph attention isomorphism neural network[C]. 2022 IEEE Intelligent Vehicles Symposium (IV), Aachen, Germany, 2022: 273–279. doi: 10.1109/IV51971.2022.9827155.
[20]	武健, 舒健生, 赵建波. 基于场论的巡航导弹作战威胁建模[J]. 四川兵工学报, 2009, 30(9): 50–52. doi: 10.3969/j.issn.1006-0707.2009.09.016. WU Jian, SHU Jiansheng, and ZHAO Jianbo. Operational threat modeling of cruise missiles based on field theory[J]. Journal of Ordnance Equipment Engineering, 2009, 30(9): 50–52. doi: 10.3969/j.issn.1006-0707.2009.09.016. (查阅网上资料,未找到本条文献英文翻译信息,请确认).
[21]	孙庆鹏, 李战武, 常一哲. 基于威力势场的多机种威胁评估方法[J]. 系统工程与电子技术, 2018, 40(9): 1993–1999. doi: 10.3969/j.issn.1001-506X.2018.09.15. SUN Qingpeng, LI Zhanwu, and CHANG Yizhe. Multi-types airplane threat assessment based on combat power field[J]. Systems Engineering and Electronics, 2018, 40(9): 1993–1999. doi: 10.3969/j.issn.1001-506X.2018.09.15.
[22]	DOUCET A, FREITAS N, and GORDON N. Sequential Monte Carlo methods in Practice[M]. New York: Springer, 2001: 582. doi: 10.1007/978-1-4757-3437-9.
[23]	BERNDT D J and CLIFFORD J. Using dynamic time warping to find patterns in time series[C]. Proceedings of the 3rd International Conference on Knowledge Discovery and Data Mining, Seattle, USA, 1994: 359–370.
[24]	HASTIE T, TIBSHIRANI R, FRIEDMAN J, et al. The Elements of Statistical Learning: Data Mining, Inference, and Prediction[M]. New York: Springer, 2009: 745. doi: 10.1007/978-0-387-84858-7.
[25]	CHUNG Junyoung, GULCEHRE C, CHO K H, et al. Empirical evaluation of gated recurrent neural networks on sequence modeling[J]. arXiv preprint arXiv: 1412.3555, 2014. doi: 10.48550/arXiv.1412.3555. (查阅网上资料,不确定本文献类型是否正确,请确认).
[26]	MISRA D. Mish: A self regularized non-monotonic neural activation function[J]. arXiv preprint arXiv: 1908.08681, 2019. doi: 10.48550/arXiv.1908.08681. (查阅网上资料,不确定本文献类型是否正确,请确认).
[27]	GLOROT X and BENGIO Y. Understanding the difficulty of training deep feedforward neural networks[C]. Proceedings of the Thirteenth International Conference on Artificial Intelligence and Statistics, Sardinia, Italy, 2010: 249–256.
[28]	姚启达, 平鹏, 朱心怡, 等. 基于机器学习的磁性元件磁芯损耗预测方法[J]. 南通大学学报: 自然科学版, 2025, 24(2): 29–38. doi: 10.12194/j.ntu.20241023001. YAO Qida, PING Peng, ZHU Xinyi, et al. Core loss prediction method for magnetic components based on machine learning[J]. Journal of Nantong University: Natural Science Edition, 2025, 24(2): 29–38. doi: 10.12194/j.ntu.20241023001.
[29]	FIGUEIRA J, GRECO S, and EHROGOTT M. Multiple Criteria Decision Analysis: State of the Art Surveys[M]. New York: Springer, 2005: 1048.
[30]	SAATY T L. Decision making with the analytic hierarchy process[J]. International Journal of Services Sciences, 2008, 1(1): 83–98. doi: 10.1504/IJSSCI.2008.017590.
[31]	SINNOTT R W. Virtues of the haversine[J]. Sky and Telescope, 1984, 68(2): 158–159.
[32]	SHI Xingjian, CHEN Zhourong, WANG Hao, et al. Convolutional LSTM network: A machine learning approach for precipitation nowcasting[C]. Proceedings of the 29th International Conference on Neural Information Processing Systems - Volume 1, Montreal, Canada, 2015: 802–810.
[33]	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, Long Beach, USA, 2017: 6000–6010.