Multi-layer Texture Detection in Robotic Additive Manufacturing of Paste-like Materials Using Attention Models of the Transformer Type in Supervised Learning

Niveau : Master 2

Période : spring 2025

Context :

Robotic additive manufacturing of earth represents a promising advancement in sustainable construction, with applications in architecture and eco-construction to reduce the environmental impact of construction processes. The surface quality of 3D-printed structures heavily depends on the textures produced, which are influenced by printing parameters and material properties. Precise control of textures is crucial for ensuring the stability and mechanical performance of the constructions.

Objective

This project aims to develop a multi-layer attention model, based on Transformer architectures, to detect and analyze textures in 3D-printed earth surfaces. The model, trained using supervised learning, will classify texture variations, thus facilitating automated quality control in robotic additive manufacturing processes.

Application

If you are motivated by industrial challenges and would like to participate in this research project, please apply by sending your CV to the following email addresses:
sebastien.levilly@univ-nantes.fr, mathieu.riand@univ-nantes.fr, raphael.polyres@univ-nantes.fr, elodie.paquet@univ-nantes.fr