Evacuation trajectory prediction of passengers in transport aircraft based on social-implicit model
-
摘要: 为精准预测运输类飞机在狭窄空间内紧急情况下的乘客疏散轨迹，构建了基于深度学习的社交隐式(Social-Implicit)模型，该模型包括社交区域、社交神经元和社交损失3个模块，通过轨迹聚类和卷积神经网络处理乘客在不同状态下的行为变化和动态交互；基于波音737-800模拟舱应急撤离试验，采集了乘客的运动状态与冲突行为数据，结合社会力模型参数标定生成了训练数据集，利用该数据集对Social-Implicit模型进行训练和验证，并从平均位移误差(ADE)、最终位移误差(FDE)、平均马氏距离(AMD)和平均最大特征值(AME)四方面评估了模型的预测效果。分析结果表明：Social-Implicit模型在构建的应急撤离数据集上表现良好，ADE为0.011 m，FDE为0.02 m，相比苏黎世联邦理工学院行人数据集和塞浦路斯大学行人数据集均降低了97%，AMD和AME精度分别提高了72.4%和94.1%，说明模型在狭窄环境中捕捉乘客轨迹变化方面表现优异；该模型的撤离时间、路径、速度和瓶颈位置预测都与社会力模型结果非常接近，疏散效率分别为1.92和1.93，路径预测能准确捕捉人员卡塞位置，速度误差不超过0.02 m·s-1，瓶颈位置误差在每平方米0.41人以内，表明模型能够模拟乘客疏散中拥堵特征；与社会力模型相比，基于深度学习的Social-Implicit模型的运行时间从65.000 s显著缩短至0.021 s，模型内存减小了78.02%，因此，Social-Implicit模型能够为民航应急疏散系统的优化提供高效、准确的轨迹预测与性能评估方法。Abstract: To accurately predict passenger evacuation trajectories in narrow spaces during emergencies of transport aircraft, a social-implicit model based on deep learning was developed. The model included three modules: social zone, social cell, and social loss. Passenger behavior changes and dynamic interactions under different conditions were processed through trajectory clustering and convolutional neural networks. The passenger movement and conflict behavior data were collected from Boeing 737-800 simulation cabin emergency evacuation experiments. These data combined with social force model parameter calibration were used to generate the training dataset. The social-implicit model was trained and validated by using the dataset, and its predictive performance was evaluated in terms of average displacement error (ADE), final displacement error (FDE), average mahalanobis distance (AMD), and average maximum eigenvalue (AME). Analysis results show that the social-implicit model performs well on the constructed emergency evacuation dataset, with an ADE of 0.011 m and an FDE of 0.02 m, representing a 97% reduction compared to the ETH BIWI walking pedestrians dataset and University of Cyprus pedestrian dataset pedestrian datasets. The accuracies of AMD and AMV improve by 72.4% and 94.1%, respectively, indicating that the model excels at capturing passenger trajectory change in narrow environment. In terms of evacuation time, path, speed, and bottleneck position prediction, the model closely aligns with the results of social force model, with evacuation efficiencies of 1.92 and 1.93, respectively. The path prediction can accurately capture the congestion points, with the speed error no exceeding 0.02 m·s-1, and the bottleneck position error within 0.41 persons per square meter, demonstrating that the model can simulate congestion characteristics during the passenger evacuation. Compared to the social force model, the deep learning model's runtime significantly reduces from 65 s to 0.021 s, and the model size reduces by 78.02%. The social-implicit model provides an efficient and accurate solution for trajectory prediction and performance evaluation in civil aviation emergency evacuation system optimization.
-
Key words:
- civil aviation /
- emergency evacuation /
- deep learning /
- trajectory prediction /
- social-implicit model
-
表 1 参试人员各项数据指标
Table 1. Various data indicators of participants
年龄/岁 人数 身高/cm 人数 体重/kg 人数 [0, 20) 4 [150, 160) 3 [40, 45) 1 [20, 25) 20 [160, 170) 19 [45, 55) 15 [25, 30) 5 [170, 180) 24 [55, 65) 18 [30, 40) 3 [180, 190) 8 [65, 75) 13 [50, 60] 22 [190, 200) 0 [75, 100) 7 
下载: 导出CSV
表 2 障碍物仿真参数
Table 2. Obstacle simulation parameters
障碍物位置 参数 座椅到主过道 A2/(m·s-2) 3.00 B2/m 0.08 主过道到客舱前部 2.00 0.50 应急出口处 3.00 0.08
下载: 导出CSV
表 3 机组人员仿真参数
Table 3. Crew simulation parameters
机组人员情况 参数 疏散人员质量/kg 60±15 期望速度/(m·s-1) 0.65±0.20 松弛时间/s 0.50
下载: 导出CSV
表 4 训练参数设置
Table 4. Training parameter settings
网络参数 结果 训练批次 128 训练轮次 100 训练帧数 8 预测帧数 12 激活函数 PReLU 学习率 0.010/0.002 相关权值 0.000 1
下载: 导出CSV
表 5 不同数据集下模型的ADE
Table 5. ADE of models under different datasets
m 数据集 城市街道 酒店入口 大学校园 商业区1 商业区2 均值 客舱 Social-Implicit[40] 0.660 0.200 0.310 0.250 0.220 0.330 0.011 S-GAN[39] 0.810 0.720 0.600 0.340 0.420 0.580 M Trajectron[40] 0.390 0.120 0.200 0.150 0.110 0.190 M ExpertTraj[44] 0.300 0.090 0.190 0.150 0.120 0.170 M Social-STGCNN[44] 0.640 0.490 0.440 0.340 0.300 0.440 0.430
下载: 导出CSV
表 6 不同数据集下模型的FDE
Table 6. FDE of models under different datasets
m 数据集 城市街道 酒店入口 大学校园 商业区1 商业区2 均值 客舱 Social-Implicit[40] 1.440 0.360 0.600 0.500 0.430 0.670 0.020 S-GAN[39] 1.520 1.610 1.260 0.690 0.840 1.180 M Trajectron[40] 0.830 0.210 0.440 0.330 0.250 0.410 M ExpertTraj[44] 0.620 0.150 0.440 0.310 0.240 0.350 M Social-STGCNN[44] 1.110 0.850 0.790 0.530 0.540 0.750 0.640
下载: 导出CSV
表 7 不同数据集下模型的AMD
Table 7. AMD of models under different datasets
数据集 城市街道 酒店入口 大学校园 商业区1 商业区2 均值 客舱 Social-Implicit[40] 3.050 0.580 1.650 1.720 1.160 1.630 1.180 S-GAN[39] 3.940 2.590 2.370 1.790 1.660 2.470 M Trajectron[40] 3.040 1.980 1.730 1.210 1.230 1.840 M ExpertTraj[44] 61.770 21.300 M 32.140 M 38.400 M Social-STGCNN[44] 3.730 1.670 3.310 1.650 11.570 2.390 4.320
下载: 导出CSV
表 8 不同数据集下模型的AME
Table 8. AME of models under different datasets
数据集 城市街道 酒店入口 大学校园 商业区1 商业区2 均值 客舱 Social-Implicit[40] 0.130 0.410 0.150 0.078 0.110 0.170 0.013 S-GAN[39] 0.373 0.384 0.440 0.355 0.254 0.361 M Trajectron[40] 0.286 0.114 0.228 0.171 0.116 0.183 M ExpertTraj[44] 0.034 0.003 M 0.005 M 0.004 M Social-STGCNN[44] 0.094 0.297 0.060 0.149 0.103 0.104 0.114
下载: 导出CSV
表 9 乘客撤离速度对比
Table 9. Comparison of evacuation speed among passengers
模型类型 平均撤离率 平均撤离速度/(m·s-1) A~C排 D~F排 G~I排 DL 0.536 0.947 0.539 0.458 SF 0.541 0.967 0.524 0.475
下载: 导出CSV
表 10 模型大小与运行时间
Table 10. Size and run time of model
模型类型 模型大小/kB 运行时间/s DL 71 0.021 SF 323 65.000
下载: 导出CSV
-
[1] Aviation Safety Network. ASN aviation safety WikiBase 1999-2021[EB/OL]. 2021, https://asn.flightsafety.org/ .[2] Airbus. A statistical analysis of commercial aviation accidents 1958-2021[R]. Blagnac: Airbus, 2022. [3] MOTEVALLI V, MONAJEMI L, RASSI M. Evaluation and mitigation of aircraft slide evacuation injuries[R]. Washington DC: Transportation Research Board, 2008. [4] MARCUS J H. A review of computer evacuation models and their data needs[R]. Washington DC: Civil Aeromedical Institute, 1994. [5] DOPART K. Aircraft evacuation testing: research and technology issues[R]. Washington DC: Office of Technology Assessment, 1993. [6] Australian Business Traveller. Supersizing the superjumbo: airbus says 1 000 seat A380 due 2020[EB/OL]. 2019, https://www.executivetraveller.com/supersizing-the-superjumbo-1-000-seat-airbus-a380-due-by-2020 .[7] 封文春, 李伟, 张桐, 等. 基于社会属性的民机复杂群体应急撤离仿真与试验研究[J]. 西北工业大学学报, 2022, 40(4): 853-864.FENG Wen-chun, LI Wei, ZHANG Tong, et al. Emergency evacuation simulation and test research of civil aircraft complex groups based on the social attribute[J]. Journal of Northwestern Polytechnical University, 2022, 40(4): 853-864. (in Chinese). [8] HARRIS D, MUIR H C. Contemporary Issues in Human Factors and Aviation Safety[M]. Hampshire: Ashgate, 2005. [9] MUIR H, THOMAS L. Passenger safety and very large transportation aircraft[J]. Aircraft Engineering and Aerospace Technology, 2004, 76(5): 479-486. doi: 10.1108/00022660410555158 [10] GALEA E R, BLAKE S J, GWYNNE S, et al. The use of evacuation modelling techniques in the design of very large transport aircraft and blended wing body aircraft[J]. The Aeronautical Journal, 2003, 107(1070): 207-218. doi: 10.1017/S0001924000013270 [11] GALEA E R. Proposed methodology for the use of computer simulation to enhance aircraft evacuation certification[J]. Journal of Aircraft, 2006, 43(5): 1405-1413. doi: 10.2514/1.20937 [12] ROBBINS C R, MCKEE S. Simulating the evacuation of a commercial airliner[J]. The Aeronautical Journal, 2001, 105(1048): 323-328. doi: 10.1017/S0001924000012203 [13] XUE Z D, BLOEBAUM C. A particle swarm optimization based aircraft evacuation simulation model-VacateAir[C]//AIAA. 46th AIAA Aerospace Sciences Meeting and Exhibit. Reston: AIAA, 2008: 1-20. [14] XUE Z D, BLOEBAUM C. Experimental design based aircraft egress analysis using VacateAir-an aircraft evacuation simulation model[C]//AIAA. 12th AIAA/ISSMO Multidisciplinary Analysis and Optimization Conference. Reston: AIAA, 2008: 6061. [15] 姜星宇, 辛月兰, 刘卫铭. 基于压缩轻量化深层神经网络的车辆检测模型[J]. 信息技术, 2020, 44(7): 23-27, 32.JIANG Xing-yu, XIN Yue-lan, LIU Wei-ming. Vehicle detection model based on compressedand lightweight deep neural network[J]. Information Technology, 2020, 44(7): 23-27, 32. (in Chinese). [16] 王纪军, 靖慧, 冯曙明, 等. 基于Faster R-CNN的仓库视频监控目标检测方法研究[J]. 信息技术, 2019, 43(7): 92-96.WANG Ji-jun, JING Hui, FENG Shu-ming, et al. Object detection method of video monitoring in power warehouse based on Faster R-CNN[J]. Information Technology, 2019, 43(7): 92-96. (in Chinese). [17] GIRSHICK R. Fast R-CNN[C]//IEEE. 2015 IEEE International Conference on Computer Vision (ICCV). New York: IEEE, 2015: 1440-1448. [18] 罗靖遥, 黄征. 基于CNN分类器和卷积的目标检测[J]. 信息技术, 2017, 41(9): 101-104, 108.LUO Jing-yao, HUANG Zheng. Target detection based on CNN classifier and convolution[J]. Information Technology, 2017, 41(9): 101-104, 108. (in Chinese). [19] REN S Q, HE K M, GIRSHICK R, et al. Faster R-CNN: towards real-time object detection with region proposal networks[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2017, 39(6): 1137-1149. doi: 10.1109/TPAMI.2016.2577031 [20] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: unified, real-time object detection[C]//IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE, 2016: 779-788. [21] LIU W, ANGUELOV D, ERHAN D, et al. SSD: single shot multibox detector[C]//Springer. European Conference on Computer Vision (ECCV). Berlin: Springer, 2016: 21-37. [22] IANDOLA F N, HAN S, MOSKEWICZ M W, et al. SqueezeNet: AlexNet-level accuracy with 50x fewer parameters and <0.5 MB model size[C]//ICLR. Proceedings of the 5th International Conference on Learning Representations. New York: ICLR, 2016: 1-13. [23] SANDLER M, HOWARD A, ZHU M L, et al. MobileNetV2: inverted residuals and linear bottlenecks[C]//IEEE. 2018 IEEE/CFV Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 4510-4520. [24] 宣琦, 李甫宪. 基于多时长特征融合的人体行为识别方法[J]. 浙江工业大学学报, 2020, 48(4): 361-366.XUAN Qi, LI Fu-xian. Action recognition method based on multiple duration feature fusion[J]. Journal of Zhejiang University of Technology, 2020, 48(4): 361-366. (in Chinese). [25] HINTON G E, OSINDERO S, TEH Y W. A fast learning algorithm for deep belief nets[J]. Neural Computation, 2006, 18(7): 1527-1554. doi: 10.1162/neco.2006.18.7.1527 [26] 谭紫阳, 高忠文, 邓宇. 基于改进极限学习机和深度神经网络融合的车辆轨迹长期预测[J]. 汽车技术, 2020(11): 16-20.TAN Zi-yang, GAO Zhong-wen, DENG Yu, et al. The long term prediction of vehicle trajectory based on improved extreme learning machine and depth neural network fusion[J]. Automobile Technology, 2020(11): 16-20. (in Chinese). [27] GUPTA A, JOHNSON J, LI F F, et al. Social GAN: socially acceptable trajectories with generative adversarial networks[C]//IEEE. 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2018: 2255-2264. [28] SADEGHIAN A, KOSARAJU V, SADEGHIAN A, et al. SoPhie: an attentive GAN for predicting paths compliant to social and physical constraints[C]//IEEE. 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. New York: IEEE, 2019: 1349-1358. [29] CARTLIDGE J, GONG S H, BAI R B, et al. Spatio-temporal prediction of shopping behaviours using taxi trajectory data[C]// IEEE. 2018 IEEE 3rd International Conference on Big Data Analysis. New York: IEEE, 2018: 112-116. [30] 乔少杰, 金琨, 韩楠, 等. 一种基于高斯混合模型的轨迹预测算法[J]. 软件学报, 2015, 26(5): 1048-1063.QIAO Shao-jie, JIN Kun, HAN Nan, et al. Trajectory prediction algorithm based on Gaussian mixture model[J]. Journal of Software, 2015, 26(5): 1048-1063. (in Chinese). [31] PECHER P, HUNTER M, FUJIMOTO R. Data-driven vehicle trajectory prediction[C]//ACM. Proceedings of the 2016 ACM SIGSIM Conference on Principles of Advanced Discrete Simulation. New York: ACM, 2016: 13-22. [32] 张凯. 基于海量出租车轨迹数据的学习与预测[D]. 西安: 长安大学, 2019.ZHANG Kai. Learning and prediction based on massive taxi trajectory data[D]. Xi'an: Chang'an University, 2019. (in Chinese). [33] MUIR H, COBBETT A. Aircraft evacuation tests—an initial assessment of the influence of various aisle configurations and lighting conditions under different evacuation scenarios[R]. Montreal: Transportation Development Centre, 1996. [34] 吴义兵. 基于社会力模型的飞机应急撤离仿真研究[D]. 天津: 中国民航大学, 2018.WU Yi-bing. Research on aircraft emergency evacuation simulation based on social force model[D]. Tianjin: Civil Aviation University of China, 2018. (in Chinese). [35] 杜月娟. 大型客机应急撤离仿真及参数影响性分析[D]. 天津: 中国民航大学, 2017.DU Yue-juan. Civil aircraft emergency evacuation simulation and analysis of parameter influence[D]. Tianjin: Civil Aviation University of China, 2017. (in Chinese). [36] 宋娜. 飞机乘员应急撤离瓶颈分析与仿真研究[D]. 天津: 中国民航大学, 2020.SONG Na. Bottleneck analysis and simulation of emergency evacuation of aircraft crew[D]. Tianjin: Civil Aviation University of China, 2020. (in Chinese). [37] PELLEGRINI S, ESS A, SCHINDLER K, et al. You'll never walk alone: modeling social behavior for multi-target tracking[C]//IEEE. 2009 IEEE 12th International Conference on Computer Vision. New York: IEEE, 2009: 261-268. [38] LERNER A, CHRYSANTHOU Y, LISHINSKI D. Crowds by example[J]. Computer Graphics Forum, 2007, 26(3): 655-664. doi: 10.1111/j.1467-8659.2007.01089.x [39] HELBING D, JOHANSSON A. Encyclopedia of Complexity and Systems Science[M]. New York: Springer, 2013. [40] MOHAMED A, ZHU D Y, VU W, et al. Social-implicit: rethinking trajectory prediction evaluation and the effectiveness of implicit maximum likelihood estimation[C]//Springer. Proceedings of the European Conference on Computer Vision. Berlin: Springer, 2022: 463-479. [41] ALAHI A, GOEL K, RAMANATHAN V, et al. Social LSTM: human trajectory prediction in crowded spaces[C]// IEEE. 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE, 2016: 961-971. [42] IVANOVIC B, PAVONE M. The trajectron: probabilistic multi-agent trajectory modeling with dynamic spatiotemporal graphs[C]//IEEE 2019 IEEE/CVF International Conference on Computer Vision (ICCV). New York: IEEE, 2019: 2375-2384. [43] MONTI A, BERTUGLI A, CALDERARA S, et al. DAG-net: double attentive graph neural network for trajectory forecasting[C]//IEEE. 2020 25th International Conference on Pattern Recognition (ICPR). New York: IEEE, 2021: 2804-2810. [44] MOHAMED A, QIAN K, ELHOSEINY M, et al. Social- STGCNN: a social spatio-temporal graph convolutional neural network for human trajectory prediction[C]//IEEE. 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR). New York: IEEE, 2020: 14424-14432. [45] ZHANG J, KLINGSCH W, SCHADSCHNEIDER A, et al. Transitions in pedestrian fundamental diagrams of straight corridors and T-junctions[J]. Journal of Statistical Mechanics: Theory and Experiment, 2011, 2011(6): P06004. -
点击查看大图
图(14) / 表(10)
计量
- 文章访问数: 10
- HTML全文浏览量: 5
- PDF下载量: 3
- 被引次数: 0
