Visual servoing of a robotic manipulator based on fuzzy logic control for handling fabric lying on a table
This paper introduces a visual servoed manipulator controller based on fuzzy logic to guide a fabric towards a sewing machine. The task of the end-effector is to handle a randomly located fabric on a table and feed it to the sewing needle along the desired seam. The proposed fuzzy controller determines the linear and angular velocity of the end-effector taking into account the current position and orientation of the fabric which derive from the vision system. The fuzzy rules are derived after studying the behavior of human workers during sewing and the membership functions are formed after simulation and extended experimentation. The experimental results demonstrate the efficiency of the system as well as the robustness of the controller performance.
Is there any form of learning incorporated in the control system for speed up the processes such as feeding a stack of fabric?
-What is the velocity of the sewing machine?
-How much time does the image processing stage take per image (frame)?
The work presented in this paper is very interesting, due to all the difficulties that are involved when handling flexible fabrics. For future work, it would be interesting to see a more robust image processing (tracking) stage since you can find fabrics of many colours which in some cases it will be similar to the colour of the table, then probably you would need to perform the segmentation to isolate the fabric using the information of the colour image which will probably demand more computational load and eventually make it more difficult to satisfy the velocity of the sewing machine condition.
Not yet. But we are working on this. Handling fabrics with curved edges is our goal for the immediate future.
This is not a real-time application yet, since the system is still under development. So far, the algorithms run in Matlab environment. At the final stage, the algorithms will be implemented on a DSP. Until then, the computing time depends on the CPU’s computing power. In our PC, the image processing stage takes about 5 sec. That’s the reason why a sewing machine has not been used yet, despite the fact that the fabric follows the desired seam line as if a sewing machine existed.
Thank you for your interesting comments. They are very useful to us, but as you can understand, they are applicable to the final stage.
We haven't used learning techniques yet. At the moment, we are working on the improvement on system performance using Genetic Algorithms in order to modify the rules and/or membership functions. Our goal is to speed up the sewing process, avoiding buckling, wrinkling and bad seam quality.
Is that possible to handle a piece of fabric with curved edjes?