Building a Robot that can Paint

I attend a monthly artist business group in Vancouver, when I told them about my idea of a painting robot the reaction was confusion “Why would I paint pictures of robots?”. I quickly learned to rephrase this idea: I actually meant that I wanted use a robot to do the action of painting.

I’ve had this idea for a long time, even more-so after taking a stone-cutting workshop with Vancouver artist Michael Binkley in 2015 and finding out first-hand that tools required to break stone are very labor intensive and exhausting and create significant dust and noise. Even finding a location is challenging as the noise created from stone cutting is not neighborly. It might sound romantic to have been neighbors in Florence to Michelangelo… but it probably was a nuisance!

Before I have my dream R&D art lab away from civilization, let alone afford a machine with the ability to cut stone, a painting robot is a much quieter project I can take on in my small studio space.

The first painting created by Alfred, I placed the bulls-eye paper where I wanted to brush to be positioned. The vision system could understand “White”, “Black” and “Clean” to dip the paintbrush in paint or clean the brush from paint.

 

Starting with existing technology

Before designing something from scratch I started with an existing robot, to understand the challenges the process of painting poses. This robot arm is Uarm from UFactory.

Part of the novelty of getting a studio robot is naming it! I named my new studio-mate Alfred (after the youngest, later joining member of the Canadian Group of Seven). Alfred is a small desktop robot that has a range of about 6” x 8”, I bought it online and it comes with open-source software. I selected this robot in particular because it already comes with a vision capabilities. Based on my experience as a mechanical engineer, I think this will be the most challenging aspect of this project.

First painting by Joanne & “Alfred”

Although Alfred did not meet my expectations of being a prolific studio mate, getting to know this robot made this whole idea even more interesting. Having worked with automotive manufacturing robots (the really expensive ones) in my past work as an R&D engineer, I quickly found out an affordable, home use desktop robot is a disappointment – but I learned so much.

Alfred’s power and accuracy were so low with it could not even overcome the friction of the marker moving while in contact with the page or go to a specific location in space. Even within the same pen stroke it would lose its z position (the vertical position from the page) – thus not making a mark or slamming into the page. I tried two strategies of controlling the brush/pen:

  1. Based on its visual position in the present moment – I used a QR code to assess the location of the brush
  2. Based on inputting G-code (same code a CNC works)

But both posed different problems that will help me design a better painting robot.

 

The accuracy of Alfred was so low that it could not repeatably find the same Z-position. This lead to marker slamming into the page, sitting too long on the page or not touching the page. Many markers were destroyed in this drawing and Alfred tipped over a few times as well.

 

Alfred struggles with straight lines and event movements in equal increments…

 

Alfred was unable to follow the G-code (typically used to program CNC machines) to mimic my signature

Ethics of a Painting Robot

I love this technical challenge – But I also love the philosophical question a painting robot poses. It creates an immediate reaction. When I talk to people about my idea the reaction is that it is not art. However, when I ask them about digital art or art created by a DSLR camera, which still requires a skilled artist to use technology, they consider that it is art. Or when I tell them of French sculptor Auguste Rodin who had assistants make the full scale statues/sculptures from his models. Yet we call him the artist.

What do you think?

 

Pin It on Pinterest

Share This