| 285 | 0 | 134 |
| 下载次数 | 被引频次 | 阅读次数 |
传统基于隐马尔科夫过程模型的航迹增强方法，受限于短记忆马尔科夫假设、高斯噪声假设等，在高机动性、强行为规律的航迹增强中表现不佳.受GPT在语言处理领域巨大成功的启发，基于Transformer架构，构建了航迹增强领域的航迹大模型T-former.通过引入Patch分割、 Mamba/Informer长序列建模、预测编码与矢量量化等技术，显著提升了T-former对航迹结构规律的学习能力.实验表明，T-former在航迹预测和滤波任务中均优于传统方法，验证了其在复杂航迹分析任务中的有效性与优越性.
Abstract：Traditional trajectory enhancement methods based on hidden Markov models(HMMs)are limited by assumptions such as the short-memory Markov property and Gaussian noise, leading to suboptimal performance in enhancing trajectories with high maneuverability and strong behavioral patterns. Inspired by the remarkable success of GPT in the field of natural language processing, we construct a trajectory foundation model named T-former for trajectory enhancement based on the Transformer architecture. By introducing techniques such as Patch segmentation, Mamba/Informer for long-sequence modeling, predictive coding, and vector quantization, T-former′s ability to learn the structural patterns of trajectories is significantly improved. Experiments demonstrate that T-former outperforms traditional methods in both trajectory prediction and filtering tasks, validating its effectiveness and superiority in complex trajectory analysis tasks.
[1]Yoon S,Lee K.Aircraft trajectory prediction with inverted transformer[J].IEEE Access,2025,13:26318-26330.DOI:10.1109/access.2025.3539742.
[2]Wu L,Wang X,Yang T,et al.M2FlightNet:multiscale modeling network for short-term flight trajectory prediction[J].IEEE Transactions on Aerospace and Electronic Systems,2010,61(5):17.
[3]Wu Y,Yu H,Du J,et al.Trajectory-BERT:Trajectory estimation based on BERT trajectory pre-training model and particle filter algorithm[J].Sensors,2023,23(22):18.DOI:10.3390/s23229120.
[4]Guo D,Wu E Q,Lin L Y.FlightBERT:Binary encoding representation for flight trajectory prediction[J].IEEE Transactions on Intelligent Transportation Systems,2023,24(2):1828-1842.DOI:10.1109/tits.2022.3219923.
[5]Guo D,Zhang Z,Yan Z,et al.A non-autoregressive multi-horizon flight trajectory prediction framework with gray code representation[J].IEEE Transactions on Intelligent Transportation Systems,2025,26(11):21069-21084.DOI:10.1109/TITS.2025.3580072.
[6]Nguyen D,Fablet R.A transformer network with sparse augmented data representation and cross entropy loss for ais-based vessel trajectory prediction[J].IEEE Access 12,2024:21596-21609.
[7]Musleh M,Mokbel M.A demonstration of KAMEL:A scalable BERT-based system for trajectory imputation[C]//Companion of the 2023 International Conference on Management of Data,2023:191-194.DOI:10.1145/3555041.3589733.
[8]Basar T.A new approach to linear filtering and prediction problems[A]//Basar T,ed.Control Theory:Twenty-Five Seminal Papers(1932-1981).New York:IEEE Press,2000:167-179.DOI:10.1109/9780470544334.ch9.
[9]Wan E A,Van Der Merwe R.The unscented Kalman filter for nonlinear estimation[C]//Proceedings of the IEEE 2000 Adaptive Systems for Signal Processing,Communications,and Control Symposium.Lake Louise,AB,Canada,2000:153-158.DOI:10.1109/ASSPCC.2000.882463.
[10]Shannon C E.Prediction and entropy of printed English[J].The Bell System Technical Journal,1951,30(1):50-64.DOI:10.1002/j.1538-7305.1951.tb01366.x.
[11]Fukushima K,Miyake S,Ito T.Neocognitron:A neural network model for a mechanism of visual pattern recognition[J].IEEE Transactions on Systems,Man,and Cybernetics,1983,SMC-13(5):826-834.DOI:10.1109/TSMC.1983.6313076.
[12]Zaremba W,Sutskever I,Hinton G E.Recurrent neural network regularization[EB/OL].arXiv:1409.2329,2014.
[13]Hochreiter S,Schmidhuber J.Long short-term memory[J].Neural Computation,1997,9(8):1735-1780.DOI:10.1162/neco.1997.9.8.1735.
[14]Vaswani A,Shardlow M,Parmar N,et al.Attention is all you need[C]//Advances in Neural Information Processing Systems 30.Red Hook:Curran Associates Inc,2017:5998-6008.
[15]Raina R,Battle A,Lee H,et al.Self-taught learning:Transfer learning from unlabeled data[C]//Proceedings of the 24th International Conference on Machine Learning.New York:ACM,2007:759-766.DOI:10.1145/1273496.1273592.
[16]Nie Y,Nguyen N H,Sinthong P,et al.A time series is worth 64 words:Long-term forecasting with transformers[EB/OL].arXiv:2211.14730,2022.
[17]Gu A,Dao T.Mamba:Linear-time sequence modeling with selective state spaces[EB/OL].arXiv:2312.00752,2023.
[18]Zhou H,Zhang S,Peng J,et al.Informer:Beyond efficient transformer for long sequence time-series forecasting[EB/OL].arXiv:2009.06871,2021.
基本信息:
DOI：10.19943/j.2095-3070.jmmia.2026.01.01
中图分类号:TP18;O211.62
引用信息:
[1]杜剑平,葛成龙,吴超逸.基于航迹大模型的航迹增强技术[J].数学建模及其应用,2026,15(01):1-9.DOI:10.19943/j.2095-3070.jmmia.2026.01.01.
2026-03-15
2026-03-15