Roillo: A social interactive robot for playrooms

1
Roillo A social interactive robot for playrooms

• Marek Michalowski
• Philipp Michel
• Selma Sabanovic
• http//www.cs.cmu.edu/marekm/projects/roillo

2
Inspired by use of social robots in autism
research

• To understand how to make robots social, look at
  what is missing when humans are not social
• Search for difference between moving toys
  (exciting) and humans (frightening)

3
Approaches

• Storytelling (Plaisant)
• Affective computing (Picard)
• Joint attention (Kozima, Scaz)
• Mobility (Michaud)
• Imitation, turn-taking (Dautenhahn)
• Human-human interaction

4
Can be explorative tools or influential devices

• Observe differences between autistic and typical
  children in human-robot social interaction
• Look for change in perception from moving object
  ? living being ? social partner
• Encourage behaviors such as eye contact and joint
  attention

5
Keepon (Kozima)

• A creature-like robot

6
Keepon (Kozima)

• A creature-like robot

7
Wed like to build a robot for such situations

• Simple, inexpensive, robust
• Attractive
• Capable of rich, varied interactions

8
A few assumptions

• Morphology should be minimally representative of
  capabilities body, head, eyes?
• Body and head movements for display of affective
  and attentional state
• Separate deictic and attentional gestures for
  understanding of shared space
• Speed and rhythmicity of motions and sounds are
  fundamental

9
Minimal morphology

• We conducted a short survey of possible designs

10
Movements of body and head for affective and
attentional state

• We would like to avoid scripting movements

11
Separate deictic from attentional gestures

• Looking at a person and pointing at an object

12
Speed and rhythmicity

• Higher-level content-rich interaction dependent
  on spatial and temporal conventions or protocols
• Spatial abilities (bodily movement, attentional
  cues, deictic gestures) must be considered
  together with temporal properties

13
Questions

• Can tasks such as requesting a ball be carried
  out more easily with separate attentional and
  deictic channels than with only one?
• Can we use rhythm and melody of natural vocal
  interactions to create a content-free dialogue?

14
Next steps

• Talking to Ben Brown about mechanical feasibility
• Talking to ETC about animation, motion,
  appearance
• Research morphology, deictic gesturing,
  rhythmicity
• Animations, surveys, storyboards, mechanical
  design?