计算机科学 ›› 2026, Vol. 53 ›› Issue (8): 117-126.doi: 10.11896/jsjkx.260700063
邓佳燕1,3, 田时瑞2, 刘厚3, 朱宁波2, 段明星3
DENG Jiayan1,3, TIAN Shirui2, LIU Hou3, ZHU Ningbo2, DUAN Mingxing3
摘要： 针对现有行人轨迹预测方法在显式运动结构建模方面的不足,以及复杂组合运动样本稀缺导致模型难以泛化至未见组合运动场景的问题,提出一种组合运动零样本行人轨迹预测网络(Compositional Motion Zero-shot Pedestrian Trajectory Prediction Network,CZP-Net)。首先,通过运动编码模块提取目标行人与邻域行人的交互运动表征,刻画目标运动趋势与邻域交互对未来轨迹的联合影响;其次,构建运动单元原型库,学习具有可组合性的运动表示,并设计原型分散约束以降低不同运动原型之间的表示冗余、增强原型判别性;随后,构造组合运动推理模块,基于交互表征与运动原型的相似度自适应生成组合权重,计算面向未见运动模式的组合运动表征;最后,设计共享运动解码器来预测多模态未来轨迹,将未来轨迹建模为运动单元组合的条件分布,并基于最大熵原理优化组合权重,使组合推理兼具概率解释性与运动语义解释性。实验结果表明,在零样本组合运动预测任务中,CZP-Net相较于最优基线模型,在平均位移误差(ADE)和最终位移误差(FDE)上分别降低47.46%和30.47%;在长尾组合运动预测任务中,ADE和FDE分别降低53.41%和45.08%。
中图分类号:
| [1] LIU Y J,JI Q G.Pedestrian trajectory prediction based on motion patterns and time-frequency domain fusion[J].Computer Science,2025,52(7):92-102. [2] WANG C D,WANG J M,GAO W B,et al.SIAT:Pedestrian trajectory prediction via social interaction-aware transformer[J].Complex & Intelligent Systems,2025,11(8):1-14. [3] GUPTA A,JOHNSON J,FEI-FEI L,et al.Social GAN:Sociallyacceptable trajectories with generative adversarial networks[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2018:2255-2264. [4] ZHANG X Y,TAN G.Real-time accurate object tracking for resource-constrained edge devices[J].Computer Science,2024,51(S2):338-346. [5] LIU H J,ZOU D P,LI P.Pedestrian trajectory prediction method based on graph attention interaction[J].Computer Science,2026,53(1):97-103. [6] GU T P,CHEN G Y,LI J L,et al.Stochastic trajectory prediction via motion indeterminacy diffusion[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2022:17092-17101. [7] GAO T C,ZHANG Y Z,GUO H,et al.SocialMP:Learning so-cial-aware motion patterns via additive fusion for pedestrian trajectory prediction[C]//Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence.2025:90-98. [8] SALZMANN T,IVANOVIC B,CHAKRAVARTY P,et al.Trajectron++:Dynamically feasible trajectory forecasting with heterogeneous data[C]//Proceedings of the European Confe-rence on Computer Vision.2020:301-317. [9] YUAN Y,WENG X S,OU Y L,et al.AgentFormer:Agent-aware transformers for socio-temporal multi-agent forecasting[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:9813-9823. [10] LIU Y,LIU Z,REN X,et al.Intention-aware diffusion model for pedestrian trajectory prediction[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2026:18469-18477. [11] PARK D,SURANA M,DESAI P,et al.Generative active lear-ning for long-tail trajectory prediction via controllable diffusion model[C]//Proceedings of the IEEE/CVF International Confe-rence on Computer Vision.2025:27839-27850. [12] BAE I,OH J,JEON H G.EigenTrajectory:Low-rank descriptors for multimodal trajectory forecasting[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2023:10017-10029. [13] FU Y X,YAN Q,WANG L L,et al.MoFlow:One-step flow matching for human trajectory forecasting via implicit maximum-likelihood-estimation-based distillation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2025:17282-17293. [14] LI R C,ZHU Z X,QIAO T Q,et al.ViTE:Virtual graph trajectory expert router for pedestrian trajectory prediction[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2026:17535-17543. [15] WANG Y,FU F,FANG M,et al.iDMaTraj:Improved diffusion Mamba model for stochastic pedestrian trajectory prediction[J].Computers,2026,15(1):12. [16] HU B,CHAM T J.TSC-Net:Prediction of pedestrian trajectories by trajectory-scene-cell classification[C]//Proceedings of the International Conference on Learning Representations.2025. [17] CHEN K,ZHAO X,HUANG Y,et al.SocialMOIF:Multi-order intention fusion for pedestrian trajectory prediction[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition.2025:22465-22475. [18] GONG S,BAO Y,HOU Y,et al.Local causal dynamic integra-ted global mode guidance transformer network for pedestrian trajectory prediction[J].PLoS One,2026,21(4):e0347049. [19] CHEN W,SANG H,WANG J,et al.DSTIGCN:Deformablespatial-temporal interaction graph convolution network for pedestrian trajectory prediction[J].IEEE Transactions on Intelligent Transportation Systems,2025,26(5):6923-6935. [20] ZHONG H,KONG Q X,CAI X Q,et al.Intelligent recognition method based on multimodal feature fusion[J].Computer Science,2026,53(S1):447-456. [21] HONG M J,JI Q G.SPP-STGCN:Spatio-temporal graph convolutional network for scene-pedestrian-pedestrian interactions[J].Computer Science,2025,52(12):133-140. [22] MANGALAM K,GIRASE H,AGARWAL S,et al.It is not the journey but the destination:Endpoint-conditioned trajectory prediction[C]//Proceedings of the European Conference on Computer Vision.2020:759-776. [23] DENG J Y,TIAN S R,LIU X L,et al.Trajectory prediction method based on multi-stage pedestrian feature mining[J].Computer Science,2025,52(9):241-248. [24] PELLEGRINI S,ESS A,SCHINDLER K,et al.You will never walk alone:Modeling social behavior for multi-target tracking[C]//Proceedings of the IEEE International Conference on Computer Vision.2009:261-268. [25] LERNER A,CHRYSANTHOU Y,LISCHINSKI D.Crowds by example[J].Computer Graphics Forum,2007,26(3):655-664. [26] ROBICQUET A,SADEGHIAN A,ALAHI A,et al.Learning social etiquette:Human trajectory understanding in crowded scenes[C]//Proceedings of the European Conference on Computer Vision.2016:549-565. |
|
||
首页