ICAART_DC 2024 Abstracts


Full Papers
Paper Nr: 5
Title:

A Deep Learning Approach to Address the Storage Location Assignment Problem

Authors:

Paul Courtin

Abstract: Automated warehouses picking performance optimization is strongly influenced by a wise assignment of products into storage locations. The storage location assignment problem (SLAP), is usually solved with exact or approach methods. For this study we focus on shuttles-based storage and retrieval system (SBS/RS). In this article we present a new method to address the SLAP using Deep Learning. We exploit historical picking order data to feed a neural network model. This model returns an optimal allocation matrix for products into the SBS/RS based on past picking orders. We present our architecture and our preliminary results on a synthetic SBS/RS.

Paper Nr: 6
Title:

Automatic Generation of Training Scenarios: Coupling Large Language Model and Ontology

Authors:

Jeanne Parisse

Abstract: The aim of this thesis is to explore ways to automatically generate training scenarios for first responders and emergency services professions for Virtual Reality training, and eventually to build a model able to do so. In particular, we will focus on professionals such as law enforcement officers, health care workers and firefighters. These scenarios will set a conflict between moral values and legal obligations in order to test the trainee in situations involving the typical vulnerabilities of their profession. We will focus on automatically generating scenarios with strong ”mise-en-scène”, i.e. storytelling visual, contextual, and other elements strengthening the emotional impact of the scenario, with the goal of inducing a stronger sense of moral conflict and emotional tension in learners that we hope will yield improved training outcomes. The key conceptual proposition underpinning this doctoral project is to evaluate the potential of coupling ontologies and generative LLMs such as OpenAI’s GPT-4. The ontology would no longer serve as the sole basis for a scenario, but would be used to elaborate an initial scenario prompt, which would then be supplied to a Large Language Model that would then create this scenario with more de- tailed scene setting elements.

Paper Nr: 7
Title:

Unsupervised Representation Learning of Complex Time Series for Maneuverability State Identification in Smart Mobility

Authors:

Thabang Lebese

Abstract: Multivariate Time Series (MTS) data capture temporal behaviors to provide invaluable insights into various physical dynamic phenomena. In smart mobility, MTS plays a crucial role in providing temporal dynamics of behaviors such as maneuver patterns, enabling early detection of anomalous behaviors while facilitating proactivity in Prognostics and Health Management (PHM). In this work, we aim to address challenges associated with modeling MTS data collected from a vehicle using sensors. Our goal is to investigate the effectiveness of two distinct unsupervised representation learning approaches in identifying maneuvering states in smart mobility. Specifically, we focus on some bivariate accelerations extracted from 2.5 years of driving, where the dataset is non-stationary, long, noisy, and completely unlabeled, making manual labeling impractical. The approaches of interest are Temporal Neighborhood Coding for Maneuvering (TNC4maneuvering) and Decoupled Local and Global Representation learner for Maneuvering (DLG4Maneuvering). The main advantage of these frameworks is that they capture transferable insights in a form of representations from the data that can be effectively applied in multiple subsequent tasks, such as time-series classification, clustering, and multi-linear regression, which are the quantitative measures and qualitative measures, including visualization of representations themselves and resulting reconstructed MVT, respectively. We compare their effectiveness, where possible, in order to gain insights into which approach is more effective in identifying maneuvering states in smart mobility.

Paper Nr: 8
Title:

An Exploratory Study on the Alignment Between Human Perception and Deep Learning Explainability Methods

Authors:

Daniel S. Costa, Pedro S. Moura and Adriana F. Alvim

Abstract: In recent years, several advances have been observed in the area of Deep Learning with surprising results. Models in this area have been increasingly used in numerous applications, including those sensitive to human life, which require clear explanations and justifications, and this has encouraged several researches into the explainability of neural networks. However, few studies have been dedicated to developing explainability methods that take human perception into account. This work proposes an investigation into the alignment of the results of explainability methods with human perception in relation to image classification.