Crystal forming robots

Christian Faubel, Crystal forming robots on overhead

The Crystal Forming Robots are little autonomous robots that are placed on an overhead projector. Each robot is powered by the light of the projector and their movements over its surface make tangible the growth process of crystal structures.

When a robot has collected enough energy, it will start moving around. The robots are equipped with tiny magnets, and as soon as two robots with matching polarity come close, they attract each other. Over time, more and more pairs of robots form, create larger clusters and a crystal like structure eventually emerges. The overhead projector magnifies the process into an abstract movie.

The background of this work are the early experiments of cybernetician Gordon Pask on building a chemical computer as a learning system. With the help of software simulation the idea of a growing structure that modifies its own perception of the environment is illustrated. The robotic implementation of the growth process is a first step towards making such a process tangible.

The robots are going to be presented in a performance and exhibition at the Sight + Sound festival in Montreal next month. The programme of the event is, as usual, rather exciting. Sadly, i can’t make it to Montreal so i figured out that the next best thing would be to talk to some of the artists who will be there. Hence this little Q&A with Christian Faubel

0s0d0clustering.jpg

Photographs of a clustering sequence, it took approximately 20 minutes for the final structure to build

Hi Christian! I’m very curious about the way the little bots move in this video. For example, what happens when they all get immobile? Is the system ‘trying to figure out’ what to do next? What controls the behavior of the robots? Why do some move and others are more passive? Is there a hierarchy?

There is no hierarchy, each of the robots is fully autonomous and triggers a movement when it has collected enough energy through its solar panel. Even though they are all built with the same components, they may have variations in timing and duration of their movement. These variations appear because the components are not perfect, they have physical differences and theses differences contribute to the behavior of the robots. Another contribution to differences in behavior, is the fact that environmental conditions on the ohp vary, in the center there is stronger light and thus more energy for the robots to harvest. As a consequence robots in the center move more often than those on the borders.

Your description of the text talks about parasites and ecosystems. The way the robots move has something a bit organic. It’s particularly uncanny in the video version with colorful umbrellas. How important is the observation or imitation of nature when you’re developing robotic artworks?

I see most of my robotic artworks as reflections on nature, I consider these robots as philosophical toys because they make the abstract concepts of autonomy and self-organisation tangible. These concepts were developed to describe and understand the way behavior is organized in living beings. So i think that ideally the artworks tell us something about ourselves.

The crystal forming robots are actually an experimental platform that i keep working on as part of my artistic research at the lab3. The first version, that is also documented in the video, had rectangular shapes, while I am currently working with hexagonal shapes. This local difference in shape has global effects in form of the growing shapes. My next step is to add contact points on the robots, so that when they cluster electrical connections are created. Once i have this in place there are so many experiments to do with growing electrical connections, i am really looking forward to this.

What is the ‘diffusion limited aggregation algorithm’, developed for simulating crystal growth? Can you explain us how it works?

The diffusion limited aggregation algorithm was developed and described in a seminal paper by Witten and Sander in the 80ties to simulate crystal growth processes. [Witten, T. t., and Sander, L. Diffusion-limited aggregation. Physical Review B 27, 9 (1983).]

The basic principle is to simulate particles that do a random walk (diffusion), when they hit a structure (by chance), they attach to that structure (aggregation). The structure is initialized with a single element, over time more and more particles dock onto the structure and a crystal like structure will form.

When you google for it you will find an overwhelming number of beautiful implementations in processing. Andy Lomas presented very nice simulations on Siggraph in 2005. I became interested in this algorithm by a general interest on growth processes and specifically through works such as Roots by Roman Kirschner, which took the works of Gordon Pask on building a chemical computer as starting point. My research on this topic is documented in a seminar on plasticity. When you scroll down you will also find some examples of experiments on crystal growth, as well as some simulations with the diffusion limited aggregation algorithm.

“Over time a crystal like structure emerges from more and more little robots forming larger clusters.” What happens once the structure has been formed? is the bots work over and done? or do they separate and start again the clustering process?

0allelelti9390.jpg

Detail of the current hexagonal prototype

No they will not separate again, the whole process runs into one direction and after an hour or more there will usually be only one single big structure. The robots need to be reset manually when the process has converged. I would say that the experiment is finished, when the process has converged and that you then start another experiment, by putting the robots apart again. What you will observe over the course of multiple experiments is that the shapes that form are always different in detail, but structurally similar.

I saw on the festival program that you will also take part in a Monochrome Layering performance at the festival. Will the Overheadbots be part of the events? Or are you going to do something that has nothing to do with them?

The overheadbots have a lot to do with the performance. When we (Tina Tonagel, Ralf Schreiber and myself) started to work on our performance project some years ago, the overheadbots were sort of a trigger for this project. In our performance, the key is the simultaneity of sound and vision. We place kinetic objects such as for example overheadbots, but also all different kind of small robots or self build instruments on the ohp and we use pick-up microphones to amplify the sound that they make when moving. So that in parallel to the moving shadow, or moving light you also hear the sound of the movement.
this is maybe best captured in these two videos:

Kunst und Musik mit dem Tageslichtprojektor @ Designacademy Eindhoven

Performance at the Shinytoys festival, September 2011

Why do you chose to work mostly with analog robots?

I like the openness of analog circuits. You don’t need to implement any sort of digital communication protocol to link up to a device. Instead you can couple thinks by simply putting a cable that creates electrical connection. For example the when the crystal forming bots are equipped with contacts, so that an electrical connection between them is created, it is enough to put that connection in between the trigger points of the two circuits and the robots will from the moment the connection is created move in synchrony. This happens without any re-programming or other re-configuration.
Another aspect of analog robots is their adaptivity to variations in the environment, that comes as an emergent property. As a matter of fact the behavior of these robots look very organic. I have explored this in more detail in a paper and presentation i gave last year at the xcoax conference in Bergamo http://2013.xcoax.org/pdf/xcoax2013-faubel.pdf.

Conceptually i like the concept of the analog, not in difference to digital computation, but estimating in contrast to counting. i have been influenced a lot by the book Analogous and Digital of the German designer and typographer Otl Aicher. In this book he writes for example that a digital clock always shows the time precisely to the second. It provides you with exact numerical values, but the landscape of time, whether it is morning or afternoon, too early or too late, i can easier deduce from the positioning of the clock hand on the clock face.

Speaking with Otl Aicher i would say that i am more interested in the landscapes than in numerical measures.

Thanks Christian!