Robotic knot tying and string manipulation

Project leads: Devin Balkcom, Weifu Wang, Matthew Bell
Funding: NSF RI 1217447, NSF CAREER (Balkcom)

Robot knot tying is worth exploring for practical reasons: we would like robots to be able to achieve tasks like helping the elderly get dressed, perform robot surgery, and bind microscale objects together for miniaturized assembly. There are also fundamental reasons to explore knot tying: robots are tradiationally good at working with simple rigid bodies, and manipulating flexible objects is a worthy research challenge.

Simplicity of design may reduce the cost and increase the reliability of robotic systems. While there is prior work using multiple robot arms and sophisticated vision systems to tie knots, we have designed simple mechanical devices with one or a few moving parts, fixtures, that quickly and reliably tie a wider variety of knots than prior approaches. One of the most interesting ideas is that devices with only four pieces can be designed to loosely arrange a knot of any complexity. A second simple device can be used to then tighten the knot. Embedding a tightening fixture inside an arrangement fixture allows a knot to be arranged and tightened without the need for transferring between fixtures. We have also explored foundational questions, including the number of fingers needed to grasp or tie a particular knot shape.