A Robotic sewing system for flexible fabric manipulation using computer vision, learning-based grasping and force control

Abstract Garment sewing remains a challenging task for robotic automation due to the highly deformable nature of fabrics and the need for coordinated perception, manipulation, and control. This paper presents an integrated robotic framework for flexible fabric manipulation and sewing, combining computer vision, learning-based grasp point estimation, motion planning, and force-regulated control within a single-arm robotic system equipped with an unmodified industrial sewing machine. The proposed approach incorporates vision-based fabric detection and contour extraction, geometry-based generation of sewing trajectories for convex and non-convex shapes and a neural network model trained on human demonstrations for grasp point estimation. Fabric positioning and sewing are achieved through a multimodal control architecture that combines vision-based feedback for seam tracking and orientation regulation with force control for tension management. Experimental evaluation was conducted on a SCARA robotic platform using fabrics of different shapes, sizes, and materials. The results demonstrate reliable fabric manipulation, accurate grasp point prediction with mean errors of only a few millimeters, and stable seam tracking along straight and curved trajectories. The measured seam deviation remained within ±2.5 mm throughout the experiments, confirming the effectiveness of the integrated perception and control framework. Although limited user intervention is still required, the results demonstrate the potential of combining vision, learning, and force control to advance robotic sewing toward more autonomous and flexible textile manufacturing applications.