Modeling and Control for Automated Sewing of Stretchable Fabrics of Different Shapes

Automating sewing remains challenging due to the deformable and condition-dependent behavior of fabrics, especially in sewing fabrics of different shapes (SFDS), where two layers of stretchable fabrics of different lengths and/or curvatures must be aligned and stitched along a desired seam. Existing robotic sewing platforms often rely on fixtures and neglect fabric deformation. This article proposes an automated sewing system that accounts for fabric deformation within planning and control. We introduce an online modeling approach based on autoregressive with exogenous input, which learns fabric sewing dynamics during operation, implicitly capturing the effect of the deformation characteristics of the fabric. Building on this model, we design a model predictive control-based controller that controls the fabric feeding direction to guide fabric layers accurately along the desired seam together with the visual servoing system. We implemented the proposed control system on an experimental prototype that comprises an overlock sewing machine, two manipulators with roller-type end effectors, a real-time sewing speed sensing module, and a seam detection module. The proposed system effectively stretches and rotates the upper and lower fabrics independently to match the seam lengths and curvatures. The experimental results demonstrate the effectiveness of the proposed scheme for SFDS.