Robot Handling Fabrics Towards Sewing Using Computational Intelligence Methods

Abstract

In cloth making industry, the need for high flexibility in automation is really imperative due to the extensive style and material variation of the products and the fast fashion changes. The automated handling of fabric materials is one of the most challenging problems in the field of robotics, since it contains a vast potential for high productivity and cost reduction. The main difficulty to get full use of this potential is the fact that it is rather impossible to predict the behaviour of fabrics towards handling due to their low bending rigidity. Moreover, it is difficult to be modelled due to their unpredictable, non-linear and complex mechanical behaviour. Sewing is the most important joining technology in garments and textiles manufacturing. The fact that sewing fabrics is a “sensitive” operation due to the erratic behaviour of fabrics poses barriers in the development of automated sewing systems. A solution to manipulation tasks that are afflicted with uncertainty, subjectivity, ambiguity and vagueness is an intelligent control approach. The development of intelligent control systems with vision feedback enables robots to perform skilful tasks in more realistic environments and make the research efforts for flexibility and automation really promising. Thus, industrial robots supported by machine vision and sensors can contribute towards advanced automation in apparel manufacturing. On the other hand, sewing fabrics with the exclusive use of robots, without human intervention, is a complicated task. The stitching process is special in the sense that the error cannot be corrected after a part of the cloth has been stitched, implying that the stitching process is not reversible. This limitation implies that clothes that do not conform to the specifications are defective and should be withdrawn from the production. This imposes a great demand on precision during the sewing process. In practice, there is a predefined tolerance considered to be acceptable, as errors resulting from robot accuracy and camera resolution cannot be eliminated. The main focus of this work lies on the design of an innovative intelligent robot controller based on visual servoing that aims to enable robots to handle flexible fabric sheets towards sewing. A great challenge is the design of a robot controller showing robustness against deformations that are likely to appear on the fabric towards robot handling. Special emphasis is given on robot handling fabrics comprised of curved edges with arbitrary curvature. This task is even harder, because up-todate industrial robots present limitations in their movement, owing to the fact that they can only be programmed to make straight or circular motions.