Robotics and computer vision for automation in the ceramic industry: design and simulation

The ceramic industry is characterized by high levels of automation in production and finished-goods storage, enabled by automated warehouses, roller conveyors, and Automated Guided Vehicles (AGVs). However, order fulfillment processes requiring partial pallet withdrawals (picking) still represent a major operational bottleneck. Opening a pallet requires removing its structural restraints, causing loss of load stability and requiring manual pallet stabilization, which increases cycle times and process variability. This thesis, developed within the industrial context of Gruppo Concorde, proposes the design and validation of an automated picking station specifically conceived for the ceramic sector. The adopted methodology integrates warehouse data analysis, mechatronic system design, robotic simulation, computer vision, and performance evaluation. Data analysis highlighted a strong concentration of picking operations on a limited number of formats and a predominance of bundle and single-box withdrawals, often requiring de-bundling operations. These results guided the definition of system requirements and the sizing of the automated solution. The proposed architecture separates the pallet opening phase from order composition through a depalletizing and de-bundling line that generates stable intermediate handling units, followed by a robotic palletizing cell for the formation of shipping units. The system was validated through an integrated simulation environment based on Unreal Engine and NVIDIA Isaac Sim, interconnected via the OPC UA protocol. Simulations verified robot reachability, collision avoidance, and PLC control logic. A computer vision was also developed for strap detection, grasp centering and crack detection. The proposed approach enables a systemic reconfiguration of the logistics flow, bridging the gap between automated storage and order fulfillment in the ceramic industry.