Human robot interaction www.biorobotics.ttu.ee Social robots - - PowerPoint PPT Presentation

www.biorobotics.ttu.ee

Human robot interaction

www.biorobotics.ttu.ee

Social robots

• Traditional robots

– Tools to perform difficult, hazardous, tedious tasks

• Social robots

– Companions and partners

www.biorobotics.ttu.ee

Social robots

• Are designed to iteract with people in a

natural manner

• Designed to achieve social-emotional goals
• Long-term goal is to create robots that are

competent and capable partners for humans

www.biorobotics.ttu.ee

Social robots

5.5

• Need a multidisciplinary approach:

– Robotics (mechanics, control theory, computer science) – Artificial intelligence – Psychology – Neuroscience – Antropology – Etc.

www.biorobotics.ttu.ee

Social robot embodiment

5.5

• Communication with humans through verbal,

nonverbal, affective modalities

– Whole body motion – Proxemics (interpersonal distance) – Gestures – Facial expressions – Gaze behaviour – Head orientation – Linguistic or emotive vocalization – Touch-based communication

www.biorobotics.ttu.ee

Antropomorphic robots - Androids

5.5

AlbertHUBO, Hanson Robotics

Zou Ren Ti

Actroid, Kokoro Company Ltd

www.biorobotics.ttu.ee

Uncanny valley

5.5

www.biorobotics.ttu.ee

Antopomorphic robots - humanoids

5.5 Sony ASIMO NEC PaPeRo Robonova, Hitec

www.biorobotics.ttu.ee

Zoomorphic robots

Paro Sony Aibo MIT Leonardo

www.biorobotics.ttu.ee

Emotion-based interaction

• Kismet: caretaker-infant pair where a

human acts as the caretaker for the robot

http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/

www.biorobotics.ttu.ee

Kismet: infant-caregiver interaction

• The infant is strongly biased to learn

communication skills that result in having the caretaker satisfy the infant's drives

• integrates perception, attention, behavior,

motivations, motor skills and expressive acts.

www.biorobotics.ttu.ee

Emotive expressions

5.5

sad happy angry sad surprise interest calm disgust

http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/

www.biorobotics.ttu.ee

Emotive responses of Kismet

http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/

www.biorobotics.ttu.ee

Emotional states

http://www.ai.mit.edu/projects/humanoid-robotics-group/kismet/

www.biorobotics.ttu.ee

RobotCub

• 45 DOF humanoid

robot

• Size of 3 years old

child

• Crawls and sits

www.biorobotics.ttu.ee

Robotcub

• visual, vestibular,

auditory, and haptic sensors

• Head and eyes

articulated

• Hands allow

complicated manipulation

http://www.robotcub.org/

www.biorobotics.ttu.ee

RoboCub Cognitive structure

• Perception
• action
• adaptation
• anticipation
• motivation

http://www.robotcub.org/

www.biorobotics.ttu.ee

Component capabilities

Compute optical flow Compute visual motion with ego-motion compensation Segmentation of the flow-field based on similarity of flow parameters Segmentation based on the presence of a temporally-persistent boundary Fixation and vergence Gaze control: smooth pursuit with prediction; possibly tuned by learning Classification of groups of entities based on low numbers Classification of groups of entities based on gross quantity Detection of mutual gaze Detection of biological motion

http://www.robotcub.org/

www.biorobotics.ttu.ee

Phylogenetic Capabilities

Saccadic re-direction of gaze towards salient multi-modal events Focus attention and direct gaze on human faces Ocular modulation of head pose to centre eye gaze Move the hand(s) towards the centre of the visual field Stabilize & integrate of saccadic percepts Stabilize gaze with respect to self-motion (VOR) Create attention-grabbing stimuli Gait control

http://www.robotcub.org/

www.biorobotics.ttu.ee

cognitive perception/ action behaviours

Object tracking through occlusion (smooth pursuit & saccades) Learn to coordinate vestibulo-ocular reflex (VOR) & tracking Learn to reach towards a fixation point Attention and action selection by modulation of capabilities Condition modulation based on anticipation Construct sensorimotor maps & cross-modal maps Learn by demonstration (crawling & constrained reaching) Exploratory, curiousity-driven, action Experience-based action selection based on interaction histories Navigate based on local landmarks and ego-centric representations

http://www.robotcub.org/