A Reconfigurable ROS2-Based Architecture for Robotic Management of Apparel and Accessories

Robotic processing of apparel items and accessories is still challenging because such products exhibit strong variability in geometry, material composition, deformability, and structure. Planar garments, multilayer textiles, footwear, and other apparel-related items require different sensing, interpretation, and execution strategies, limiting the generalization of highly task-specific solutions. This paper presents a reconfigurable ROS2-based robotic architecture for heterogeneous apparel items and accessories. The framework integrates visual perception, semantic feature extraction, coordinate transformation, task planning, motion planning, execution, and feedback handling into a unified pipeline deployable on a single-arm robotic platform. The aim is to provide a flexible, multi-task architecture for heterogeneous items and operations in realistic industrial workflows. Rather than targeting a single product category or operation, the architecture analyses the visual and geometric characteristics of the target item and configures the processing workflow accordingly. The framework is here instantiated for selective disassembly, a demanding circular economy use case requiring accurate localization and treatment of seams and heterogeneous components. A case study on jumpers and footwear demonstrates how the same architectural backbone can support substantially different item geometries and processing requirements.