This video demonstrates the progress we have thus far made with regards to robotic manipulation of paper. Working with paper affords us a very rich interaction domain that has as yet not been solved for anthropomorphic robot hands. We use a marker based visual detection mechanism that serves as input for soft-body physical modelling of the sheet of paper. Modelling of the paper was implemented using the Bullet Physics Engine. The high speed fiducial marker detection framework is freely available in our open-source computer-vision library, ICL (http://www.iclcv.org
In the first part of the video, visual detection and physical modelling of paper is demonstrated. The developed system is able to track a standard sheet of paper (with fiducial markers) while it is folded into a paper aeroplane. In the second part of the video,
the results of the detection and modelling units are used to implement a complex folding sequence carried out on our bi-manual 54-DOF Robot setup consisting of two
PA-10 Robot arms, each equipped with a 20-DOF Shadow Dexterous Hand.