Transcript The MIT Artificial Intelligence Lab

The Next Generation of Robots?
Rodney Brooks
and
Una-May O’Reilly
MIT Artificial Intelligence Laboratory — Research Directions
Our Objectives
• How can biology inform robotic
competence?
• How can aspects of human
development and social behavior
inform robotic competence?
MIT Artificial Intelligence Laboratory — Research Directions
Our Approach
• Exploit the advantages of the robot’s physical
embodiment
• Integrate multiple sensory and motor systems
to provide robust and stable behavioral
constraints
• Capitalize on social cues from an instructor
• Build adaptive systems with a developmental
progression to limit complexity
MIT Artificial Intelligence Laboratory — Research Directions
Our Humanoid Platforms
Cog
and
Kismet
MIT Artificial Intelligence Laboratory — Research Directions
Biological Inspiration for Cog
• Cog has simulated musculature in its arms
• Cog has an implementation of a human model of
visual search and attention
• Cog employs context-based attention and internal
situations influence action
• Cog uses a naïve model of physics to distinguish
animate from inanimate
MIT Artificial Intelligence Laboratory — Research Directions
Social Inspiration for Cog
• A theory of mind
• A theory of body
• Mimicry
MIT Artificial Intelligence Laboratory — Research Directions
Human <—> Robot
Cameras
Gaze direction
Eye tilt
Facial
features
Speech synthesizer
Microphones
Right eye pan
Camera with
wide field of
view
Camera with
narrow field of
view
Axis of rotation
Left eye pan
Neck tilt
Neck pan
Neck lean
Head orientation
MIT Artificial Intelligence Laboratory — Research Directions
Levels of Control
Social Level
robot responds
to human
Behavior Level
perceptual
feedback
Skills Level
coordination
between motor
modalities
Primitives Level
human responds
to robot
current goal
current primitive(s)
MIT Artificial Intelligence Laboratory — Research Directions
Kismet’s Competencies
• Direct Visual Attention
• Recognize Socially Communicated
Reinforcement
• Communicate Internal State to Human
• Regulation of Social Interaction
MIT Artificial Intelligence Laboratory — Research Directions
No One in Charge
Cameras
Eye, neck, jaw motors
Ear, eyebrow, eyelid,
lip motors
QNX
Motor
ctrl
Attn
system
•
Eye
finder
CORBA
Tracker
Speakers
NT
Skin
filter
speech synthesis
affect recognition
•
Dist.
to
target
Motion
filter
Color
filter
audio
speech
comms
L
dual-port
RAM
Face
Control
Percept
& Motor
Emotive
Response
Drives &
Behavior
•
• 11 400-500 MHz PCs
CORBA
Linux
Speech
recognition
Microphone
CORBA
– QNX (vision)
– Linux (speech
recognition)
– NT (speech
synthesis & vocal
affect recognition
MIT Artificial Intelligence Laboratory — Research Directions
4 Motorola
68332 microcontrollers
L, multithreaded lisp
higher-level
perception,
motivation,
behavior,
motor skill
integration &
face control
Visual Attention
Frame Grabber
skin tone
color
motion
w
w
w
Top down,
task-driven
influences
reset
w
inhibit
attention
habituation
Eye Motor
Control
MIT Artificial Intelligence Laboratory — Research Directions
Visual Search
MIT Artificial Intelligence Laboratory — Research Directions
Social Constraints
Person
backs off
Too close –
withdrawal
response
Person draws
closer
Comfortable
interaction distance
Too far –
calling
behavior
Beyond
sensor
range
Comfortable interaction
speed
Too fast,
Too close –
threat response
Too fast –
irritation
response
MIT Artificial Intelligence Laboratory — Research Directions
Cross Cultural Affect
That’s a good bo-o-y!
pitch, f (kHz)
o
pitch, fo (kHz)
No no baby.
time (ms)
time (ms)
approval
MMMM Oh, honey.
pitch, f (kHz)
o
Can you
get it?
pitch, f (kHz)
o
Can you
get it?
prohibition
time (ms)
attention
time (ms)
comfort
Evidence for 4 contours in
Kismet-directed speech
MIT Artificial Intelligence Laboratory — Research Directions
Affect Recognizer
energy variance
Soothing &
Low-Intensity
neutral
vs
Everything Else
Soothing
vs
Low-Intensity Neutral
Approval & Attention
vs
Prohibition
vs
High Intensity Neutral
prohibition
soothing
approval
vs
attention
neutral
approval
attention
prohibition
neutral
attention & approval
pitch mean soothing & low-energy neutral
MIT Artificial Intelligence Laboratory — Research Directions
Naive Subjects
•
5 female subjects
• 4 naive subjects
• 1 caregiver
•
Four contours and neutral
speech
•
•
Multiple languages
•
•
praise, prohibition, attention,
soothing
French, German, Indonesian,
English, Russian
Driven by Human
MIT Artificial Intelligence Laboratory — Research Directions
Facial Expressions
arousal
angry
surprise
afraid
stress
elated
excitement
frustrated
displeasure
pleasure
neutral
sad
happy
content
depression
fatigued
calm
relaxed
bored
sleepy
sleep
MIT Artificial Intelligence Laboratory — Research Directions
Facial Postures in Affect Space
Open
stance
Low
arousal
fear
accepting
Negative
valence
tired
unhappy
content
surprise
Positive
valence
disgust
High
arousal
stern
anger
Closed stance
MIT Artificial Intelligence Laboratory — Research Directions
Face, Voice, Posture
MIT Artificial Intelligence Laboratory — Research Directions
Turn-Taking / Proto-Dialog
•Naïve subjects
•Told to “talk to the robot”
•Engage in turn taking
•No understanding (on either side) of content
MIT Artificial Intelligence Laboratory — Research Directions
Implemented Model of Visual Search and Attention
Color
w
Motion
w

Motor
System
Activation
Map
Skin
w
Habituation
w
Motivation
System
MIT Artificial Intelligence Laboratory — Research Directions
MIT Artificial Intelligence Laboratory — Research Directions
Hardware – Cog’s Arms
Encoder
Bearing
Motor and
Gearbox
Actuator
output
Spring
•
•
•
•
6 DOF in each arm
Series elastic actuator
Force control
Spring law
F ( x)  kx  bx
MIT Artificial Intelligence Laboratory — Research Directions
Hardware – Cog’s Head
• 7 degrees of freedom
• Human speed and range of motion
MIT Artificial Intelligence Laboratory — Research Directions
Visual and Inertial Sensors
3-axis inertial sensor
Peripheral View
Peripheral View
Foveal View
Foveal View
MIT Artificial Intelligence Laboratory — Research Directions
Computational System
• Designed for real-time
responses
• Network of 24 PC’s
ranging from 200-800 MHz
• QNX real-time operating
system
• Implementation shown
today consists of
– ~26 QNX processes
– ~75 QNX threads
MIT Artificial Intelligence Laboratory — Research Directions
MIT Artificial Intelligence Laboratory — Research Directions