Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios

Kai Chen; Xiao Song; Hang Yu

doi:10.1007/978-981-15-1078-6_3

Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios

Kai Chen
, Xiao Song^*
, Hang Yu

^*Corresponding author for this work

School of Cyber Science and Technology

Beihang University
Tsinghua University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

3 Scopus citations

Abstract

Pedestrian trajectory prediction is a challenging problem in the crowded and chaotic scenarios. Currently, the prediction error is still high because the input of Long Short-Term Memory (LSTM) network is a 1D vector, which cannot represent the spatial information of pedestrians. To tackle this, we propose to use tensors to represent the complex environmental information. Meanwhile, LSTM internal full connection is converted into full convolution to predict the spatiotemporal pedestrian trajectory sequences. The results show that our method reduces the displacement offset error better than recent works including Social-LSTM, SS-LSTM, CNN, Social-GAN, Scene-LSTM, providing more realistic trajectory prediction for the chaotic crowd.

Original language	English
Title of host publication	Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings
Editors	Gary Tan, Yong Meng Teo, Axel Lehmann, Wentong Cai
Publisher	Springer
Pages	29-39
Number of pages	11
ISBN (Print)	9789811510779
DOIs	https://doi.org/10.1007/978-981-15-1078-6_3
State	Published - 2019
Event	19th Asia Simulation Conference, AsiaSim 2019 - Singapore, Singapore Duration: 30 Oct 2019 → 1 Nov 2019

Publication series

Name	Communications in Computer and Information Science
Volume	1094
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	19th Asia Simulation Conference, AsiaSim 2019
Country/Territory	Singapore
City	Singapore
Period	30/10/19 → 1/11/19

Keywords

Convolutional neural network
Pedestrian behavior
Trajectory prediction

Access to Document

10.1007/978-981-15-1078-6_3

Cite this

Chen, K., Song, X., & Yu, H. (2019). Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios. In G. Tan, Y. M. Teo, A. Lehmann, & W. Cai (Eds.), Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings (pp. 29-39). (Communications in Computer and Information Science; Vol. 1094). Springer. https://doi.org/10.1007/978-981-15-1078-6_3

Chen, Kai ; Song, Xiao ; Yu, Hang. / Conv-LSTM : Pedestrian Trajectory Prediction in Crowded Scenarios. Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings. editor / Gary Tan ; Yong Meng Teo ; Axel Lehmann ; Wentong Cai. Springer, 2019. pp. 29-39 (Communications in Computer and Information Science).

@inproceedings{81e8edf2fa2943e98f6a038487bba7a4,

title = "Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios",

abstract = "Pedestrian trajectory prediction is a challenging problem in the crowded and chaotic scenarios. Currently, the prediction error is still high because the input of Long Short-Term Memory (LSTM) network is a 1D vector, which cannot represent the spatial information of pedestrians. To tackle this, we propose to use tensors to represent the complex environmental information. Meanwhile, LSTM internal full connection is converted into full convolution to predict the spatiotemporal pedestrian trajectory sequences. The results show that our method reduces the displacement offset error better than recent works including Social-LSTM, SS-LSTM, CNN, Social-GAN, Scene-LSTM, providing more realistic trajectory prediction for the chaotic crowd.",

keywords = "Convolutional neural network, Pedestrian behavior, Trajectory prediction",

author = "Kai Chen and Xiao Song and Hang Yu",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Singapore Pte Ltd.; 19th Asia Simulation Conference, AsiaSim 2019 ; Conference date: 30-10-2019 Through 01-11-2019",

year = "2019",

doi = "10.1007/978-981-15-1078-6\_3",

language = "英语",

isbn = "9789811510779",

series = "Communications in Computer and Information Science",

publisher = "Springer",

pages = "29--39",

editor = "Gary Tan and Teo, \{Yong Meng\} and Axel Lehmann and Wentong Cai",

booktitle = "Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings",

address = "德国",

}

Chen, K, Song, X & Yu, H 2019, Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios. in G Tan, YM Teo, A Lehmann & W Cai (eds), Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings. Communications in Computer and Information Science, vol. 1094, Springer, pp. 29-39, 19th Asia Simulation Conference, AsiaSim 2019, Singapore, Singapore, 30/10/19. https://doi.org/10.1007/978-981-15-1078-6_3

Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios. / Chen, Kai; Song, Xiao; Yu, Hang.
Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings. ed. / Gary Tan; Yong Meng Teo; Axel Lehmann; Wentong Cai. Springer, 2019. p. 29-39 (Communications in Computer and Information Science; Vol. 1094).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Conv-LSTM

T2 - 19th Asia Simulation Conference, AsiaSim 2019

AU - Chen, Kai

AU - Song, Xiao

AU - Yu, Hang

PY - 2019

Y1 - 2019

N2 - Pedestrian trajectory prediction is a challenging problem in the crowded and chaotic scenarios. Currently, the prediction error is still high because the input of Long Short-Term Memory (LSTM) network is a 1D vector, which cannot represent the spatial information of pedestrians. To tackle this, we propose to use tensors to represent the complex environmental information. Meanwhile, LSTM internal full connection is converted into full convolution to predict the spatiotemporal pedestrian trajectory sequences. The results show that our method reduces the displacement offset error better than recent works including Social-LSTM, SS-LSTM, CNN, Social-GAN, Scene-LSTM, providing more realistic trajectory prediction for the chaotic crowd.

AB - Pedestrian trajectory prediction is a challenging problem in the crowded and chaotic scenarios. Currently, the prediction error is still high because the input of Long Short-Term Memory (LSTM) network is a 1D vector, which cannot represent the spatial information of pedestrians. To tackle this, we propose to use tensors to represent the complex environmental information. Meanwhile, LSTM internal full connection is converted into full convolution to predict the spatiotemporal pedestrian trajectory sequences. The results show that our method reduces the displacement offset error better than recent works including Social-LSTM, SS-LSTM, CNN, Social-GAN, Scene-LSTM, providing more realistic trajectory prediction for the chaotic crowd.

KW - Convolutional neural network

KW - Pedestrian behavior

KW - Trajectory prediction

UR - https://www.scopus.com/pages/publications/85075679805

U2 - 10.1007/978-981-15-1078-6_3

DO - 10.1007/978-981-15-1078-6_3

M3 - 会议稿件

AN - SCOPUS:85075679805

SN - 9789811510779

T3 - Communications in Computer and Information Science

SP - 29

EP - 39

BT - Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings

A2 - Tan, Gary

A2 - Teo, Yong Meng

A2 - Lehmann, Axel

A2 - Cai, Wentong

PB - Springer

Y2 - 30 October 2019 through 1 November 2019

ER -

Chen K, Song X, Yu H. Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios. In Tan G, Teo YM, Lehmann A, Cai W, editors, Methods and Applications for Modeling and Simulation of Complex Systems - 19th Asia Simulation Conference, AsiaSim 2019, Proceedings. Springer. 2019. p. 29-39. (Communications in Computer and Information Science). doi: 10.1007/978-981-15-1078-6_3

Conv-LSTM: Pedestrian Trajectory Prediction in Crowded Scenarios

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this