Robotic knitting

The project "Do robots dream about knitting? that was initiated in 2018. Re-encoding the cooperation between robots and humans" at the University of Berlin researches new forms of the human/machine interaction by teaching a robot to knit and thus also breaking away from traditional sex stereotypes. We held an interview with the Project Manager, Dr. phil. Patricia Treusch on this exciting topic.

How does one actually come up with the idea of starting a research project that teaches robots how to knit? What do you want to achieve or find out?

The idea arose when I was working as a technology sociologist in a robotics laboratory where new robotic arms were created and the question was being posed: Which scenarios of interaction do we want to use them for? At the time I suggested as a joke and somewhat spontaneously "knitting". At the first glance it seems funny, combining a "cutting edge" technology with the seemingly "most boring cultural technology in the world". But the starting point was however the question: "How do we want to work together with the robotic arms of the latest generation that were made for the physically close work together with 'us people'. While searching for photos online I often came across two photos: On the one hand the photo of the highly-automated, flexible robotic arm working together with a male craftsman. On the other hand, the arm of a robot carrying out care services together with a female carer. In both photos the fact that the photos work using certain social stereotypes stands out.

It is my aim to generate new photos of people and robots working together and thus bringing hands, grippers and activities together in a different way without conjuring up the really cliché-like already existing images. Running around a robotics laboratory with knitting needles and wool to bring needles, thread, grippers and hands together to enable a joint knitting procedure goes against the established pattern of thinking.

What is the status of your project "Do robots dream about knitting? Recoding the collaboration between robots and people"? Are there initial cognitions?

The process of cooperative knitting is the focus of the project. We defined three scenarios here that we wanted to realise:

1. A person knitting with a robotic arm handing him the thread. 2. A person knitting with a robotic arm taking over the (empty) right needle 3. Two robotic arms taking over the two needles and the person handing him the thread or helping orchestrate the thread, needles and stitches.
It quickly became clear how complicated knitting actually is and how many compensatory movements I actually make as the knitter. My task here is to translate knitting into a learnable movement for the robot. Finally, we have also ascertained that the right material proves to be a big challenge. For instance, we used needles in different thickness made of plastic, bamboo, and metal, but also a wide range of different wool qualities and thicknesses in order to find out what is conducive for the process of joint knitting. In the meantime, we can quite robustly hand one of the needles to PANDA and carry out scenario 2.

Do computers learn how to knit in a different way to people?

Absolutely. For instance I learned how to knit by watching YouTube video tutorials for socks. PANDA learns how to knit by following the knitting movements I make moving along the points of the x, y and z axes. PANDA follows the pattern moving from point to point, holding the robotic joints properly so that the desired knitting movement occurs. PANDA doesn't know that we are knitting nor that there is a thread that are steering.

The project "Do robots dream about knitting? that was initiated in 2018. Re-encoding the cooperation between robots and humans" at the University of Berlin researches new forms of the human/machine interaction by teaching a robot to knit and thus also breaking away from traditional sex stereotypes. We held an interview with the Project Manager, Dr. phil. Patricia Treusch on this exciting topic.

How does one actually come up with the idea of starting a research project that teaches robots how to knit? What do you want to achieve or find out?

The idea arose when I was working as a technology sociologist in a robotics laboratory where new robotic arms were created and the question was being posed: Which scenarios of interaction do we want to use them for? At the time I suggested as a joke and somewhat spontaneously "knitting". At the first glance it seems funny, combining a "cutting edge" technology with the seemingly "most boring cultural technology in the world". But the starting point was however the question: "How do we want to work together with the robotic arms of the latest generation that were made for the physically close work together with 'us people'. While searching for photos online I often came across two photos: On the one hand the photo of the highly-automated, flexible robotic arm working together with a male craftsman. On the other hand, the arm of a robot carrying out care services together with a female carer. In both photos the fact that the photos work using certain social stereotypes stands out.

It is my aim to generate new photos of people and robots working together and thus bringing hands, grippers and activities together in a different way without conjuring up the really cliché-like already existing images. Running around a robotics laboratory with knitting needles and wool to bring needles, thread, grippers and hands together to enable a joint knitting procedure goes against the established pattern of thinking.

What is the status of your project "Do robots dream about knitting? Recoding the collaboration between robots and people"? Are there initial cognitions?

The process of cooperative knitting is the focus of the project. We defined three scenarios here that we wanted to realise:

1. A person knitting with a robotic arm handing him the thread. 2. A person knitting with a robotic arm taking over the (empty) right needle 3. Two robotic arms taking over the two needles and the person handing him the thread or helping orchestrate the thread, needles and stitches.

It quickly became clear how complicated knitting actually is and how many compensatory movements I actually make as the knitter. My task here is to translate knitting into a learnable movement for the robot. Finally, we have also ascertained that the right material proves to be a big challenge. For instance, we used needles in different thickness made of plastic, bamboo, and metal, but also a wide range of different wool qualities and thicknesses in order to find out what is conducive for the process of joint knitting. In the meantime, we can quite robustly hand one of the needles to PANDA and carry out scenario 2.

Do computers learn how to knit in a different way to people?

Absolutely. For instance I learned how to knit by watching YouTube video tutorials for socks. PANDA learns how to knit by following the knitting movements I make moving along the points of the x, y and z axes. PANDA follows the pattern moving from point to point, holding the robotic joints properly so that the desired knitting movement occurs. PANDA doesn't know that we are knitting nor that there is a thread that are steering.