RSS 2017 Workshop.

 

Experimenting with Movement Observation

 

In brief:  This workshop is a follow-on to last year’s workshop “Embodied Experience and Movement Observation for Roboticists” where multiple somatic and choreographic techniques were explored.  This year, we will be zooming in on experimental challenges posed by this way of working — with expert, trained human movers capable of many, many complex modes of behavior.  Expect an interactive, high-energy workshop!

Organizers:  Amy LaViers alaviers@illinois.edu, Kayhan Ozcimder ozcimder@princeton.edu

Date and Time:  July 15 9am-noon

Abstract: There is a growing interest in literature to study bio-inspired multi-agent robotic systems that involve individual level interactions in order to achieve a desired shared task. In human agents, these interactions are both verbal and nonverbal, and this poses unique challenges for moving machines, which will inherently be communicating along this nonverbal channel in human-facing scenarios (intentionally or not). Thus, a major subset of these studies explore the language used by human agents that guides the creation of expressive movement phrases in collective motion as well as communication. This tutorial/workshop hybrid addresses some of the challenges faced in these explorations with an interactive two track discussions. In the first track, we will begin the discussion by proposing a new framework, dance, to study movement and collective behavior. We will show that dance is an accessible medium to formally study decision-making and communication strategies of individuals in a group and the rules extracted from dance can be used to construct new formal methods of cooperation and decision-making for multi-agent robotic systems.  In the second track, we will `zoom in’ to discuss how to modulate the individual level interactions through movement by proposing, Labanotation, a rich notational system for transcribing and analyzing movement. In particular, we will show how Laban/Bartenieff Movement Studies (LBMS) provides a taxonomy as well as a field of trained practitioners who are experts in developing motion profiles with specific meaning and intention.  Finally, the session will conclude with an interactive activity that merges these two approaches.

Description and Event Schedule:

30min – Intro + Meet and Greet (with a movement twist)

45min – Tutorial in Experimental Design

45min – Tutorial in Movement Observation

1 hour – Joint mini-experiment workshop (two tracks)

Intended audience: Curiosity on the part of roboticists into dance and the arts is a natural thing – both groups of people practice the art of movement design. This is why I believe the workshop will have strong interest from the RSS community.  Researchers in HRI, HMI, and human-in-the-loop control may see the most direct application to their work because of the “human” element.  In these fields experimental set up and interface design involve more subjective decision making and require compatibility with a human population.  However, researchers in robot learning also make often subjective decisions for example when choosing which basic movements to seed their learning algorithms with.  This workshop may help them think more carefully about this choice.  Even researchers in pure control theory and hardware design can benefit a lot from more nuanced understanding of how people have dissected the craft of movement design.

 

[ For more detail read below! ]

 

Experimental Design Tutorial Description (Kayhan Ozcimder): 

A universal setup in studying multi-agent robotic models is through the lessons learned from existing biological systems and by exploring their strategies for collective behavior. There are many domains of such biological models involving cooperation of individuals who also have to  make decisions in order to strike a balance between their individual preferences and the group objectives. There are, for instance, honey bees using waggle-dance to communicate a food source, or insects (such as ants) cooperating for navigation and foraging. These explorations have guided scientists and engineers to construct new formal methods of cooperation and decision-making for multi-agent robotic systems. However, one major constraint is the difficulties with experimentation. This is due to the challenges experienced when training insects or animals to accomplish tasks in laboratory environment or controlling the conditions when experimenting in their natural habitat.

 

In this tutorial, I will propose a new framework, dance, to study movement and human collective behavior. I will explain the design processes to create experimental setups in collaboration with dancers and choreographers to address challenging engineering problems. Moreover, I will highlight the criticality of generating a bidirectional research study. In one direction, I will show how to apply the rules and principles learned from performing arts to inspire ideas and construct formalisms that can be implemented onto robotic platforms. In the reverse direction, I will discuss how to create mechanisms to further understand human decision-making in a complex task environment, such as creating an artistic piece. During the discussion, I will refer to the existing work along this line, as well as my own studies that investigate salsa dance, and an improvisational piece, as examples of such explorations. I will demonstrate and teach elementary level dance moves with a mathematical perception of communication-through-motion, leader-follower interactions as well as group behavior.

 

Movement Observation Tutorial Description (Amy LaViers):

As we work to bring robots out of the factory and into humans’ everyday life, it is important to begin designing the expression of their movement with greater care.  In human work places, public spaces, homes, and even bodies, the pattern of movement which each platform produces will become essential to understand and to design intentionally.  Indeed, it seems that in natural human settings the movement of humans encodes information. This information may deal with environmental state (‘is the building burning?’), task state (‘how thick is the paint?’), or emotional state (‘are you in a hurry?’ and even ‘are you upset?’).  How does one find a ground truth for such communications in an experimental setup?  An area of work that is promising for helping in this task is the performing arts where professionals are trained in techniques for design and description of movement.  In particular, Laban/Bartenieff Movement Studies (LBMS) provides a taxonomy as well as a field of trained practitioners who are experts in developing motion profiles with specific meaning and intention. 

 

In this tutorial participants will explore movement from an embodied perspective.  Participants will learn about LBMS, a set of concepts, terms, and principles utilized by many dancers and actors, in three short sessions led by a roboticist trained in this work.  The goal of this session will be to attune participants to the philosophy of many movement professionals and provide key takeaways that may be of use in their research.  Synergistically, we will investigate the practice of movement observation.  Participants will engage in creating and observing movement phrases using this taxonomy.  A key point of emphasis will be around how large the space of human movement expression and perception is, motivating the need for broad movement characterizations which may help guide experimental set ups and research practices.

 

Joint Workshop Description:

Building on the work presented by Amy and Kayhan in their presentations, we will run a mini experiment to see how multi-agent interactions can be modulated through expressive movement. Without any prescribed rules, together with participants we will design an experimental platform that uses human movement to study collective behavior.  The goal will be generating a framework that integrates concepts such as communication-through-motion and leader-follower interactions borrowed from the first track to the use of Labanotation as a rich notational system for transcribing and analyzing movement.  This will be open-ended, interactive, and fun!