Nathan Melenbrink — How to Make [almost] Anything

Final Project Proposal

The final project proposal is a human-computer interactive painting device, or "Hyper-Paint-by-Number". Physically, it involves a 2 or 3 axis robotic aparatus which would be mounted to a wall or free standing. The end effector would be likely a marker or other means of making outlines on the canvas. The machine would invite passersby to participate in filling colors onto the canvas. The computer might make suggestions as to what colors to use, though the participants are of course free to ignore this, paint outside of the lines, and so on. The computer must then adapt to this haptic input, continuously redrawing outlines overtop of the user input. This would require a camera to be mounted on the end effector, constantly sending image updates back to a base computer, which would then employ computer vision methods to interpret the colors added by participants.

Tools: Milling machine, metal working tools, circuitry and electronics tools.
Hardware: Aluminum frames, actuators, canvas, camera, end effector (paint brush, marker, etc.
Software: Custom interface involving CV color tracking, blob recognition, etc. as input and outputting new tooling paths in realtime
Year: 2014

The above proposal was a perhaps overly ambitious initial idea. Over the semester the idea has evolved to the above rendered wire plotting robot. I became more interested in the idea of a portable, deployable robot that would be somewhat scalable and able to calibrate itself. This concept seemed to point towards wire robotics. More details on my final page.

This is not meant to suggest a preconceived aesthetic...it could be as abstract or photorealistic as need be. Perhaps the machine keeps striving for photorealism while the capriciousness of the decisions of individuals makes this impossible.

There are a number of reference projects dealing with robotic reproduction of art, including this one by Alex Kiessling, in which a number of robots in different cities simultaneously replicate his motions in real-time. However, this is simply replication, and does not deal with the possibilities of feedback loops or any kind of interaction.

The system would require at least two actuators for the X and Y directions, and potentially a third to control the Z dimension, which would change the pressure of the end effector on the canvas.