Vision aided neuro-fuzzy control for the folding of fabric sheets

This paper presents a new manipulator end-effector motion-planning controller, based on fuzzy logic and machine vision, for the folding of fabric sheets. The folding process is analyzed and is divided into three stages, the initial laying, the true folding and the final laying. For each stage, the handling parameters and path domains are identified through experiments. Then, a neuro-fuzzy system is formulated for the off-line motion-planning in each stage and is trained via back-propagation, utilizing the data acquired through experiments. An on-line controller with vision feedback is combined with the neuro-fuzzy to deal with the necessary modifications of the initial path. Two cameras are properly positioned to identify the position of the edges of the fabric, necessary for the vision feedback. The proposed neuro-fuzzy sub-system is then verified for different fabric materials not used for its training.