High density polyurethane blocks handling with an AI-powered multimodal vision system on a custom compact omnidirectional mobile robot: a case study

The Mobile Robot System with Artificial Intelligence (MOROSAI) project focuses on the development of a prototypical solution based on an Automated Guided Vehicle capable of autonomously moving semi-finished materials via the clever use of different smart devices and sensors. The case study involves moving large polyurethane foam blocks on a production line. Experimental activities were carried out using RGB data collected from an innovative sensor in this operational context. The analysis employed a Vision Transformer model, namely DINOv2, to explore the effectiveness of foundational models as an unsupervised learning strategy on the acquired data. The results highlight the model’s capability to extract and discriminate relevant visual features across a wide range of realistic scenarios, suggesting its potential for real-time employment in autonomous vehicle applications.