Thesis defense "Unsupervised deep learning: Application to the detection of abnormal situations in the autonomous train environment".

Summary

This thesis addresses the challenges of environmental monitoring and anomaly detection, particularly of obstacles, for an autonomous freight train. Although rail transport has traditionally been under human supervision, autonomous trains offer potential advantages in terms of cost, time and safety. Nevertheless, their operation in complex environments poses significant safety challenges. Instead of a supervised approach requiring expensive and limited annotated data, this research adopts an unsupervised technique, using unlabeled data to detect anomalies by relying on techniques capable of identifying atypical behavior.

Two environmental monitoring models are presented: the first, based on a convolutional autoencoder (CAE), is dedicated to the identification of obstacles on the main track; the second, an advanced version incorporating the vision transformer (ViT), focuses on general environmental monitoring. Both exploit unsupervised learning techniques for anomaly detection.

The results show that the method put forward provides relevant elements for monitoring the environment of the autonomous freight train, having potential to enhance its reliability and safety. The use of unsupervised techniques thus demonstrates the usefulness and relevance of their adoption in an application context for the autonomous train.

Keywords

artificial intelligence, deep learning, autonomous train, anomaly detection, environmental monitoring, autonomous freight train, rail transport, supervised approach, unsupervised approach, learning algorithms, vision transformer, convolution.

Jury composition

Mr YASSINE RUICHEK, Professeur des universités, Université de technologie de Belfort Montbéliard (Rapporteur)
M. NOREDDINE ABGHOUR, Professeur des universités, Université Hassan II Maroc (Rapporteur)
Ms LAETITIA JOURDAN, University Professor, University of Lille (Examiner)
M. SMAIL NIAR, Professeur des universités, Université Polytechnique Hauts-de-France (Thesis co-director)
M. ABDELMALIK TALEB-AHMED, Professeur des universités, Université Polytechnique Hauts-de-France (Thesis co-director)
M. LOTFI ABDI, Research Engineer, IRT Railenium (Co-supervisor)