Education, Science, Technology, Innovation and Life
Open Access
Sign In

Trajectory Generation Method Based on Deep Learning

Download as PDF

DOI: 10.23977/jeis.2024.090302 | Downloads: 16 | Views: 491

Author(s)

Jiaxiang Gao 1, Xiangjie He 1, Yiwei Liao 1

Affiliation(s)

1 Institute of Computer Science and Information Engineering, Harbin, 150025, China

Corresponding Author

Jiaxiang Gao

ABSTRACT

In recent years, trajectory data publishing has brought great convenience to our daily life, however, directly publishing real trajectory data can cause serious threats to users' privacy. In this paper, we focus on the trajectory generation problem, aiming at generating trajectory datasets similar to real trajectories to meet the demand for trajectory data for urban autopilot simulation and traffic analysis tasks, and at the same time, protect the privacy of users' trajectories. We propose a trajectory generation scheme incorporating a Variational Auto-Encoder (VAE), which is capable of generating trajectory data that is highly similar to the real trajectories, to replace the user's real sensitive trajectory data for the purpose of trajectory privacy protection. We test the proposed scheme in terms of trajectory similarity, and the results show that the proposed scheme can generate trajectory datasets more accurately and stably, and at the same time protect the user's trajectory privacy.

KEYWORDS

Variational Self-Encoder, Trajectory Data, Privacy Protection

CITE THIS PAPER

Jiaxiang Gao, Xiangjie He, Yiwei Liao, Trajectory Generation Method Based on Deep Learning. Journal of Electronics and Information Science (2024) Vol. 9: 9-13. DOI: http://dx.doi.org/10.23977/10.23977/jeis.2024.090302.

REFERENCES

[1] Daraio E, Cagliero L, Chiusano S, et al. Comple⁃ menting location-based social network data with mobility data: a pattern-based approach [J]. IEEE Transactions on Intelligent Transportation Systems, 2022, 23(11): 212-227. 
[2] Srivastava R K, Koutník J, et al. LSTM: A search space odyssey [J]. IEEE transactions on neural networks and learning systems, 2016, 28(10): 2222-2232.
[3] Zhang H, Huangfu W, Hu X, et al. Inferring the Most Popular Route Based on Ant Colony Optimization with Trajectory Data[C]//Wireless Sensor Networks: 11th China Wireless Sensor Network Conference, CWSN 2017,12-14.
[4] Jiang H, Li J, Zhao P, et al. Location privacy-preserving mechanisms in location-based services: A comprehensive survey[J]. ACM Computing Surveys (CSUR), 2021, 54(1): 1-36.
[5] S. Semeniuta, A. Severyn, E. Barth, et al. A hybrid convolutional variational autoencoder for text generation, EMNLP 2017 (2017) 627–637.
[6] Ivanovic B, Leung K, Schmerling E, et al. Multimodal deep generative models for trajectory prediction: A conditional variational autoencoder approach [J]. IEEE Robotics and Automation Letters, 2020, 6(2): 295-302.
[7] A. Gupta, J. Johnson, L. Fei-Fei, et al. Social GAN: socially acceptable trajectories with generative adversarial networks, CVPR (2018) 2255–2264.
[8] Li Y, Pan Q, Wang S, et al. Disentangled variational auto-encoder for semi-supervised learning[J]. Information Sciences, 2019, 482: 73-85.

Downloads: 10606
Visits: 361369

Sponsors, Associates, and Links


All published work is licensed under a Creative Commons Attribution 4.0 International License.

Copyright © 2016 - 2031 Clausius Scientific Press Inc. All Rights Reserved.