Design for Collaborative Robotic Environments
CfD conversation 2022–03 | April 1st, 2022
Kristian Kloeckl, Associate Professor in the Department of Art + Design and the School of Architecture, and Director of the Experience Design Lab here at Northeastern curated a panel of academics working within robotics. Each panelist gave a glimpse of their collaborative work with robots; they shared case studies, provocations, and the theoretical underpinnings of their work.
In introducing the panelists, Kristian gave his hope for the future of human-robot collaboration. His work in robotics is centered around industrial environments, where he is investigating how robots can be responsive to human needs and operations. Instead of machines placing their modalities on workers, machines should be part of responsive systems which adapt to human needs. This may sound great in theory, but it leaves us with a set of epistemological implications:
- How should machines work if they adapt to human needs?
- How do workers want to work?
- Who makes decisions in this collaborative process?
- How do we design for robots & humans to understand each other?
The panelists each had their own unique view of human-robot collaboration, but as was pointed out during the discussion portion of the event — robotics doesn’t exist yet as a discipline, so everyone at the proverbial table sits at the same distance from it, meaning that there is no one expert. Each discipline at the table is just as integral and necessary as the others.
Robots-in-the-loop
Daniel Cardoso, Associate Professor, Co-Director CodeLab, School of Architecture, Carnegie Mellon University, presented “Rethinking Automation in Construction: a Reinforcement Learning Framework for Robotically Supported Cooperative Construction.” A few key words to highlight here: robotically supported and cooperative.
Daniel and his team are designing new human-machine workflows. He describes his work using the term ‘Robot-in-the-loop’ which is a play on the more commonly used ‘human-in-the-loop’ (HITL). HITL is a branch of Artificial Intelligence where the human is involved in training and testing algorithms — Daniel’s robot-in-the-loop, on the other hand, uses reinforcement machine learning. Reinforcement learning (RL) is a type of machine learning that allows for trial and error in the formation of real-time, step-wise decision-making.
Construction sites are dynamic in nature. Reinforcement learning’s step-wise decision-making is crucial to robotically-supported cooperative construction because it allows the robots to respond dynamically to situations. RL can also be applied to robots’ social navigation as seen in the above GIF demonstrating the spatial negotiation of movement in a group of robots.
Daniel also stressed the importance of acknowledging expertise that already exists, in this case, the knowledge already contained within construction teams and trade workers. It is for this reason that Daniel follows the robot-in-the-loop model, with his emphasis on the words robotically-supported and cooperative. Because construction does not and should not be reinvented in order to adopt the use of robotic assistance, robots should support the way construction is currently being done by experts in the field.
5 Human Dispositions
Keith Evan Green, Professor, Director Architectural Robotics Lab, Department of Design + Environmental Analysis, Sibley School of Mechanical and Aerospace Engineering, Cornell University, started off his presentation by asking: where is our home as humans in an increasingly digital world? To which he responds “robots for living in,” borrowing the term from Bill Mitchell, MIT Professor and Director of the Media Lab’s Smart Cities Group. This is architectural robotics.
So then, why do we as humans need architectural robotics? Keith’s answer to this comes in the form of five human dispositions:
- A Psychological Disposition —Human beings are protean. We seek “many and new ‘places’ to exist.” (1)
- An Evolutionary Dispotision — We exhibit “adaptive flexibility,” the capacity to “buffer and adjust to environmental dynamics.” (2)
- A Learning Disposition — “The brain wants variation. It wants to move.” (3) Changing environments create new associations in our brains.
- A Poetic Disposition — We have a “burning passion for the coming-into-being of things” — a “framework we can consolidate, always differently.” (4)
- An Emotive Disposition — Machines can have a light side.
Keith then gave a set of fundamental questions for architectural robotics, some in line with those Kristian started the panel with: How much intelligence? What kind of human-robot relationship? What kind of human-robot interaction? How figurative should our perception of the robot be? How rigid should our robots be?
Keith closed his presentation by appealing to the poetic human disposition, prompting the audience to see a future where we exist in “Ecosystems of bits, bites, and biology.” He then recited lines from a poem…
of a cybernetic meadow
where mammals and computers
live together in mutually
programming harmony— “All Watched Over By Machines Of Loving Grace” by Richard Brautigan
4 (inseparable) elements of Experiential Robotics
Taskin Padir, Associate Professor, College of Engineering and Director of the Institute for Experiential Robotics, Northeastern University introduced us to Experiential Robotics by presenting its four inseparable elements. These elements of experiential robotics are foundational to Northeastern University’s new Institute for Experiential Robotics.
- Human Experiences — It’s not about robots, it’s about humans and the experiences that robots afford: from aging in place, to improving infrastructure, to helping solve society’s complex problems, and beyond.
- Learning Robots — robots should learn from their own experiences. What happens when a robot is placed in a novel environment or with novel objects and tools? How does it know how to respond?
- Robots learning from their interactions — with their collaborators and companions. The personalization of an interaction is key because an interaction, even with the same human, can vary greatly.
- Get researchers out of the lab! — Researchers need to be equipped with the experiences that they are researching and/or designing interventions for.
For the question of why experiential robotics, Taskin listed: Future-of-work, future manufacturing, first- and last-mile ecosystems, workforce shortages, technology shift to remote work, human wellbeing, and customer promise.
Taskin also discussed an NSF project that he and Kristian are working on together called Co-worker Robots to Impact Seafood Processing (CRISP). Right now in the US, there is a $20 billion trade deficit in seafood not because we don't have the fish supply, but because we simply don’t have the workforce to process it. The human element that the robots are being designed for here is a more ergonomic experience of processing large, heavy, and hard-to-move fish. The researchers have visited seafood factory floors multiple times over the course of the project thus far, as seen in the picture with both Taskin and Kristian dressed head-to-toe in seafood processing gear. The team recognizes the importance of understanding the seafood factory working experience in order to design robotic interventions for it.
Conversation takeaways
The presentations were followed by an evocative conversation on disciplines in robotics and a beautiful point made by Keith that we could all benefit from hearing:
“[..]respecting your own discipline and finding the value of that in collaborative teams, as opposed to thinking you have to cover all the ground on your own.”
— Keith Evan Green
- The human-to-robot relationship likely shouldn’t be 1:1. Daniel Cardoso even goes as far as avoiding the word “collaboration” when speaking of humans and robots as it implies a symmetry of human to robot…and as Taskin’s 1st element of Experiential Robotics states, it should really be all about the human experience.
- Robotics is an emerging field where interdisciplinary thought and action are required — No single engineering field can claim robotics as its own.
- Everyone has knowledge that can be applied in robotics, and everyone has (or should have) a seat at the table, even dancers… as we already learned in a previous CfD Conversation, but Keith pointed to here as well. Read Keith’s research in HCI with contemporary dancers here.
- And so, we need to make sure designers have a seat at the robotics table.
The questions that I’m left contemplating now are — was the table made in a robotically-supported construction environment? Can the table stand if it's missing one of its four (inseparable) legs? And finally, could our tables of the future be robots-for-conversing-around? 🙃
References
(1) Borrowed from Robert Jay Lifton in The Protean Self: Human Resilience in the Age of Fragmentation (1994).
(2) Anton, Potts, and Aiello, “Evolution of Early Homo: An Integrated Biological Perspective,” Science (2014)
(3) Benedict Care, How we learn: The Surprising Truth About When, Where, and Why it Happens. (2014)
(4) F.T. Marinetti, “The Birth of a Futurist Aesthetic,” (1911).
Interested in learning more? Watch the event recording
CAMD Moderator
Kristian Kloeckl, Associate Professor, Director Experience Design Lab, Department of Design, School of Architecture, CAMD, Northeastern University
Panelists
Daniel Cardoso Llach, Associate Professor, Co-Director CodeLab, School of Architecture, Carnegie Mellon University
Keith Evan Green, Professor, Director Architectural Robotics Lab, Department of Design + Environmental Analysis, Sibley School of Mechanical and Aerospace Engineering, Cornell University
Taskin Padir, Associate Professor, Director Institute for Experiential Robotics, College of Engineering, Northeastern University
Join us! #CenterForDesign
️📩 centerfordesign@northeastern.edu
🔗 Linkedin 🔗Twitter 🔗Facebook 🔗Website🔗Newsletter