Wednesday, October 27, 2010

Beanbag robotics





Here’s a neat idea that I’ve written up for my Material Witness column in the November issue of Nature Materials.

It’s a commonplace observation in robotic engineering that some of the hardest tasks for robots are the ones we do without thinking: balancing upright, say, or catching a ball. Even the simple feat of picking up objects, when considered as a problem in control systems engineering, becomes a formidable challenge. How should we position the fingers on approach, where should we grip the object, how much pressure should we apply? Answering these questions generally requires exquisite feedback between vision, motor control, and tactile sensing, not to mention (in our case) a fair degree of intuition and training.

The ingenuity that has gone into solving these problems in robotics is exhilarating, as exemplified by the very recent reports in this journal of pressure-sensing ‘smart skin’ [1,2]. But these solutions tend to be predicated on the assumption that a robotic hand will follow the human prototype in having several gripping fingers. The widespread use of this design in the animal world testifies to its virtues, but there’s no escaping the demands it makes on actuation, sensing and feedback.

Now Eric Brown of the University of Chicago and his coworkers have described a new design for a robotic gripper that dispenses altogether with these difficulties by replacing active control with passive adaptability. Their device has no fingers at all, but instead uses a soft mass that moulds itself to the shape of the object to be gripped [3]. The crucial aspect of the design is that, once configured in this way simply by pressing onto the object, the gripper undergoes a transition from soft to hard, becoming a rigid body encasing enough of the object to hold it with, in general, an appreciable force.

That is achieved by filling the body of the gripper – an elastic latex bag – with granular material, such as tiny glass spheres or, in one prototype, ground coffee. Rigidification of the conformable grainy mass is then induced by evacuating the air between the grains, causing slight compaction. This is sufficient to trigger a jamming transition: the grains enter a collective state of immobility, like that in a blocked funnel, which, as Brown’s coauthor Heinrich Jaeger explains in another preprint [4], is a non-equilibrium state directly analogous to a glass. Indeed, such a packing-induced transition between solidity and fluidity is familiar to anyone who has ever opened a vacuum-packed packet of coffee.

Once rigid, the gripper holds an object by a combination of three mechanisms: friction, suction caused by deformation of the jammed bag as it lifts, and geometrical ‘wrap-around’ interlocking. The resultant gripping force depends on the geometry of the object, but a whole variety of forms, from steel springs to raw eggs, can be securely held. What is more, the device works in the wet, and can grip several different objects at once while retaining their orientation. Much as in the case of walking robots [5], it shows how smart use of passive control can greatly simplify the engineering problem.

References


1. Takei, K. et al., Nat. Mater. 9, 821-826 (2010).
2. Mannsfeld, S. C. B. et al., Nat. Mater. 9, 859-864 (2010).
3. Brown, E. et al., preprint http://www.arxiv.org/1009.4444.
4. Jaeger, H. & Liu, A. J., preprint http://www.arxiv.org/1009.4874.
5. Collins, S. H.,Wisse, M., Ruina, A. & Tedrake, R., Science 307, 1082-1085 (2005).

No comments: