Artificial Intelligence and Robotics

Todd Bryant
Sareen Engineer
Han Hu
Presentation Overview
 Definition of robotics
 Robotics’ relevance to AI
 Current developments in the field
 Current implementations
 Past successes in robotics
 Roadblocks to robotics research
 Future of robotics
Definition of Robotics
 A robot is…

– “An active artificial agent whose environment is

the physical world”
--Russell and Norvig

– “A programmable, multifunction manipulator

designed to move material, parts, tools or
specific devices through variable programmed
motions for the performance of a variety of
tasks”
--Robot Institute of America
Relevance to Artificial Intelligence

 Effectors
 Sensors
 Architecture
 Integration of various inputs
– Hierarchy of information representation
 Emotions
Effectors
 Effector vs. Actuator
 Degrees of freedom
(d.f.)
– 6 d.f. for free body in
space
 Locomotion
– Statically stable vs.
Dynamically stable
 Manipulation
– Rotary vs. Prismatic Four-finger Utah/MIT hand
motion
– End Effector
Sensors

 Force-sensing
 Tactile-sensing
 Sonar
 Visual (camera)
 Proprioceptive

Robot with camera

attached
Architecture
 Classical architecture
– shortcomings
 Behavior-based architecture

Reason about behavior of objects

Identify objects
Sensors Actuators
Build maps

Avoid objects

Design for a behavior-based mobile robot

(adapted from Fig 25.10 in AIMA)
Information Representation Hierarchy
 Raw data
 Cognitive
feature
 Conceptual
feature
 Simple
concept
 Inter-
connected
synthesized
concept
Information Representation Hierarchy
Information Representation Hierarchy
Current Developments
 Emotions
 Energy-efficiency
 Integration
– Hierarchy of information representation
 Control structures
– Synthesis of neural nets and fuzzy logic
 Robotic surgery
– Telepresence
 Robot perception
– Face and object recognition
Importance of Emotions
 Emotions help prevent people from
repeating their mistakes (decisions that
resulted in negative feelings)
 Recognizing emotions would allow
robots to become more responsive to
users’ needs
 Exhibiting emotions would help robots
interact with humans
Classification of Emotions
 Continuous
– Emotions defined in multi-dimensional
space of attributes
– Arousal-Valence Plane

 Discrete
– Defines 5, 6, or more “basic” emotional
states upon which more complex
emotions are based
Arousal-Valence Plane
 Valence – whether emotion is positive or
negative
 Arousal – intensity of emotion
Classification of Emotions
Plutchik’s Theory:
– Eight primitive emotions that more complex
emotions are based upon
• Gladness (joy)
• Sadness
• Anger
• Surprise
• Acceptance
• Disgust
• Expectancy
• Fear
Complexity of Emotional Classification
Affective Research: Kismet
 Decides proper
emotional response to
stimuli and exhibits
corresponding facial
expression, body
posture, and vocal
quality
 Behavioral response
serves either social or
self-maintenance
functions
Kismet smiling
Organization of Kismet’s Emotions
 Some of Kismet’s emotions, what causes
those emotions, and what purpose they serve
Kismet
Energy-Efficiency: Seaglider
 Small electric pump
transfers 100cm3 of
oil from an external
bladder to its
reservoir, making
Seaglider dense
enough to sink
 To dive, small motor
pushes battery pack
into nose
 Process is reversed
to ascend
Seaglider’s diving process
Current Implementation
 Industrial robots
– used in factories to
manufacture boxes and
pack and wrap
merchandise
 Car manufacturers own
50% of today’s robots
 Robots used in
hazardous situations
– Nuclear power plants
– Response to bomb threat
– Outer space exploration Robotic arm arranging
chocolates
Current Implementation: Asimo

Honda’s Asimo (Advanced

Step in Innovative Mobility)
 Able to walk freely (can

change stride speed)

 Able to balance on one

foot
 Able to climb stairs

 Able to manipulate objects

 Space- & cost-efficient

Honda’s Asimo
 Asimo’s Recognition Technology
 Based on visual cues
such as the angle and
distance at which it
perceives an object
 Can map an object's
contour and compare
it to a database of
prototypes for
different expressions
and actions
 Is currently limited to ASIMO making measurements
pre-registered people
Successes: AIBO
 1996 – Prototype with
small body, camera,
microphones, and
batteries
– Trouble balancing
– Required further
development of
software
 First generation AIBO
ERS-110 released in
Japan (1999)
AIBO

• Example of robotics SPECIFICS

with AI  64-bit RISC processor
• Behavior dependent  18 joints
on owner’s behavior  Touch sensor
– No two are alike  CCD color camera
• Voice recognition  Infrared distance sensor
– 50 distinct  Acceleration sensor
commands  Angular velocity sensor
Cog – Brain
 No central unit  Core network – 200 MHz
 Heterogeneous network PC’s running QNX real-
of different processors time OS, connected by
 Microcontrollers (such 100VG Ethernet
as Motorolla 6811) – 4 nodes
process inputs and
drive motor responses
 A/V processing done by
digital signal processor
(DSP) networks
 Relay data to core
processor network by
way of ISA and PCI
cards
Cog – Sensory Systems
 Visual System  Vestibular System
– Binocular – 3 semi-circular
– Each eye has 2 gray canals mimicked by 3
scale cameras rate gyroscopes
 Auditory System – 2 linear
accelerometers
– 2 microphones
– Stereo sampling @
 Tactile System
22.05 kHz with 8-bit – 6x4 array of sensors
resolution on torso can detect
– Sound localization position and force of
has been achieved, a touch
currently working on – Some implementation
segregation of sound in hands
streams
Roadblocks In Research:
Interactive Activity
Roadblocks In Research
 Shift attention from manufacturing to
design processes
 Shift attention from single to multiple
capabilities
 Energy-related issues
– Bulky batteries with short lifespan
Problems

 Sensing
 Vision
 Mobility
 Design
 Control
 Reasoning
Problems
 Sensing  Vision
– Cost of tactile – Two methods
sensors very high • Corner recognition
– Range Limits • Edge recognition
• Light – 2 meters – Overlap of objects
• Required(factory) – 10 – Visibility of local
meters features
Problems
 Mobility  Design
– Growing need for – Control of robot after
AGV’s in outdoor construction
applications – Development of
– Vision and laser knuckles required to
ranging systems perform such tasks
need development to as lifting and
produce information grasping well
at a faster rate – Actuators are often
– Current bipeds are too big, slow, or
incapable of walking difficult to control
on uneven ground
Problems
 Control  Reasoning
– Simulation is not – AI (an essential
accurate to real world component of
interaction robotics) has slowly
– Based on been introduced into
mathematical and industrial world
numerical – Further refinement in
computations this field before
faster progress of
robotics
Future of Robotics
 Downsizing  Greater intelligence
– Reduction in power – More user-friendly
needs and size interface

 Synergism  More
environmentally
– Greater integration of
technologies
friendly
– Robots easy to
disassemble and
destroy
– Easily reusable or
degradable parts
Future of Robotics
 Design robots to
recognize presence,
posture, and gaze
 Develop viable
social exchange
between robots and
humans
 Design systems that
can learn via
reinforcement
Moral Dilemmas
 Legal rights of
autonomous beings
 Replacing humans in
the workplace
 Ethics of deleting
intelligent robots—
murder?
 Creating helpful
sentient robots vs.
playing God
Any Questions?