Please use this identifier to cite or link to this item:
Title: Pedestrian motion prediction using deep generative networks
Authors: Ong, Xing Long
Keywords: DRNTU::Engineering::Electrical and electronic engineering
Issue Date: 2019
Abstract: For a mobile robot to assist humans in daily life like attending to a patient in a hospital by serving him/her meals, it is crucial for the robot to be able to recognise the motion behaviour of a people so that the robot can create an intelligence guess to avoid colliding with the people. Besides, human navigation behaviour may be influenced by the surrounding people and static obstacles in the vicinity. Various methods have been presented over the years for understanding motion behaviour such as social force model and a recurrent neural network (RNN). Undeniably, the new state-of-art method would be deep learning which became the most popular research topic. It can handle complex situations and able to handle a vast amount of data and to learn deep features automatically. This project objective aims to explore pedestrian motion prediction using Generative Adversarial Network (GAN), a recurrent sequence-to-sequence model by observing past pedestrian motion history and predict their future location. The process of training a GAN is to train a recurrent discriminator to discriminate between acceptable and fake motion by continuously feeding in data. The trained model is then used to generate future location of the pedestrian. Therefore, the robot can utilize this information and make decision in advance such as colliding with humans even in a crowded environment. Before implementation on robots, it is mandatory to test the GAN model with open source pedestrian datasets. From the results gathered from our GAN model, it is visible to conclude that GAN can predict the future motion of the pedestrians for safe and efficient trajectory planning. It is then compared with the traditional methods of using a linear model, Kalman Filter in future pedestrian trajectory prediction.
Rights: Nanyang Technological University
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Student Reports (FYP/IA/PA/PI)

Files in This Item:
File Description SizeFormat 
  Restricted Access
Pedestrian Motion Prediction Using Deep Generative Networks4.16 MBAdobe PDFView/Open

Page view(s) 50

checked on Oct 20, 2020

Download(s) 50

checked on Oct 20, 2020

Google ScholarTM


Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.