计算机科学 ›› 2026, Vol. 53 ›› Issue (6A): 250600118-8.doi: 10.11896/jsjkx.250600118
张晓涵1, 杨飞2, 马靖尧1, 赵韩栎1, 赵旭3
ZHANG Xiaohan1, YANG Fei2, MA Jingyao1, ZHAO Hanyue1, ZHAO Xu3
摘要： 针对复杂环境下单一模型轨迹预测精度不足的问题,提出一种基于改进状态空间模型与记忆网络融合的新型预测架构通道注意力增强型双流记忆网络CA-MLNet。该工作的核心创新在于:(1)通过结构化状态空间模型(Structured State-Space Model,SSM),增强对动态环境的空间特征选择能力;(2)创新性地融合xLSTM的mLSTM模块作为辅助时序建模单元,构建双分支特征融合架构。改进后的SSM模块能有效捕捉长距离空间依赖,而mLSTM模块通过指数门控机制强化局部时序特征提取,两者通过自适应权重融合机制实现优势互补。在GeoLife GPS Trajectories行人轨迹数据集上的实验表明,所提模型在预测场景下达到99.01%的预测精度,较基准模型提升8.18%,相比xLSTM架构提升3.14%。消融实验验证了SSM模块改进对空间特征选择准确率的贡献达43.6%,双模块协同工作使轨迹偏移误差降低,为智能交通预警系统提供了高精度的解决方案。
中图分类号:
| [1] LIU W X,LI Z,YAN H W.Electronic Fence Intrusion Detection Technology Based on YOLOv5 Algorithm[J].Industrial Control Computer,2024,37(7):94-95,98. [2] HUO X M,YU Y.Application Research of Maritime “SmallTraffic Control” Based on Electronic Fence Technology[J].Navigation,2025(1):38-41. [3] WIEST J,HÖFFKEN M,KREβEL U,et al.Probabilistic trajectory prediction with Gaussian mixture models[C]//2012 IEEE Intelligent Vehicles Symposium.Madrid:IEEE,2012:141-146. [4] HELLBACH S,EGGERT J P,KÖRNER E,et al.Time seriesanalysis for long term prediction of human movement trajectories[C]//International Conference on Neural Information Processing.Berlin:Springer,2008:567-574. [5] NIKHIL N,MORRIS B T.Convolutional neural network fortrajectory prediction[C]//European Conference on Computer Vision Workshops.Munich:Springer,2018:1-12. [6] XIE G,SHANGGUAN A,FEI R,et al.Motion trajectory prediction based on a CNN-LSTM sequential model[J].Science China Information Sciences,2020,63(11):212201. [7] BECK M,PÖPPEL K,SPANRING M,et al.mLSTM:Extended Long Short-Term Memory[J/OL].https://arxiv.org/abs/2405.04517. [8] GU A,DAO T.Mamba:Linear-time sequence modeling with selective state spaces[EB/OL].arXiv,2023.https://arxiv.org/abs/2312.00752 [9] GU A,GUPTA A,GOEL K,et al.Efficiently modeling long sequences with structured state spaces[J].arXiv:2111.00396,2021. [10] CHAMAN M,MALIKI E A,YANBOIY E H,et al.Comparative analysis of deep neural networks YOLOv11 and YOLOv12 for real-time vehicle detection in autonomous vehicles[J].International Journal on Transport Development and Integration,2025,9(1):45-53. [11] DELORD M,DOUIRI A.Multiple states clustering analysis(MSCA),an unsupervised approach to multiple time-to-event electronic health records applied to multimorbidity associated with myocardial infarction[J].BMC Medical Research Methodo-logy,2025,25(1):32. [12] LI H R,YU H X,HE Y X.HPnet:Hybrid Parallel Network for Human Pose Estimation[J].Sensors,2023,23(9):4321. [13] ZHANG H,WANG Y,CHEN L.DynaScale:Dynamic kernelgeneration for adaptive multi-scale attention in video understanding[J].IEEE Transactions on Multimedia,2023,29(4):1001-1015. [14] WANG Q,LI X.T-Mamba:Temporal mamba with spatiotemporal cooperative attention for traffic flow prediction[J].Neural Networks,2024,178:123-136. [15] XIE G.Principles of GPS and Receiver Design[M].Beijing:Electronic Industry Press,2017:436. [16] SCHMIDHUBER J.A fixed-size storage O(n) time complexity learning algorithm for sequence prediction[J].Neural Computation,1991,3(4):490-501. [17] XU X,YAO W Q,LI C,et al.Modulated Signal Classification and Recognition Algorithm Based on Gated Attention Network[J].Modern Electronics Technique,2025,48(3):69-75. [18] LI F L,SHI H,YI K F,et al.Detecting Navigable Areas of Unstructured Roads by Fusing Road Direction and Multi-Feature Information[J/OL].Journal of Computer-Aided Design & Computer Graphics,2025.http://kns.cnki.net/kcms/detail/11.2925.tp.20250227.0916.002.html. [19] CAI N B,ZHANG C,WANG W.Multi-Step Rolling Prediction Method Based on Long Short-Term Memory(LSTM) Neural Network Model[J].Science Technology and Engineering,2024,24(19):8356-8361. |
|
||
首页