It was important to keep the butterflies light and delicate–ideally as delicate as an actual butterfly–while allowing the wings to move in response to stimulus from the environment. We ultimately decided on laser-cut, heavyweight drawing paper to make the wings themselves, and we devised a mechanism using neodymium magnets and a servo for the wing movement.

Each wing was laser cut from the drawing paper, then fitted with 2 pairs of small neodymium magnets, making sure that the polarity on all magnets was kept consistent. Under the mounting board sat a much larger magnet oriented with the opposite polarity of the magnets on the wings. For example if the magnets on the wings all had the south pole facing down, the large magnet had south facing up. A hobby servo was used to rotate the large magnet so as it moved closer to the wings, the magnets on the wings were repelled causing them to close. All the servos were connected to a single computer using an Arduino board which was in turn connected to mac mini running a processing application which used a webcam mounted at the back of the room to look for movement in front of each butterfly. As visitors walked closer to get a better view, the butterfly in the pedestal they approached begins to flap its wings.

While working on the mechanism for the physical butterflies, we also started working on possible methods for patterning the wings. Andrew and robert had been working on a generative system based on simulated charged particles for some prints in another room of the gallery; and they modified some of these systems to generate point distributions that were in turn used to create voronoi diagrams fitted to the wing profile. The voronoi patterning eventually produced two of the families of butterflies. In addition to the “straight” voronoi, another version was created that fit b-spline curves to the voronoi cells.

Another set of patterns were produced using a fluid simulation and vector field visualization algorithm. The wing profiles were used as boundaries for fluid flow systems. This field was then fixed and rendered using fitted lines and circles.

The fifth pattern family was created by drawing random lines in the wing profile and placing circles at the point of intersection. The lines and intersections created cells that were then offset to create a structure for the wing.

Inspired by Benjamin Aranda and Chris Lasch’s book Tooling. I developed a variation of the cracking algorithm found in that book which introduced b-spline fitting to smooth the finaly output and give a more organic feel. The algorithm “cracks” a polygon by dividing it from and edge to its centroid, each resultant polygon is the put in a pool to be chosen at random and recursively cracked again.

The final patterning system was a simple random circle packing algorithm. Randomly sized circles were placed in the wing shape, if there was any overlap with the already existing circles, the new one was tossed and the process run again until the wing was filled. This algorithm produced a surprising degree of variation by altering the sensitivity of the overlap testing function.

A few members of each family were printed and mounted on the wall using insect pins, giving the whole thing and extra classic feel. The final movement turned out to be very creepy. Having them react to activity, not proximity, really helped add to the overall feel. As one approached a case it almost felt like as if they were flapping in hopes of escape, only to give up a few moments later.

Here is a quick video showing the movement and the space. You can see the mechanism in there too.

You can see some more photos on flickr.