Transcript Contents and Teams: - City University of New York

All Teams Overview:
Team:
Topics
1
•Overview
•Chapter 1: From Teleoperation to Autonomy
•Chapter 2: The Hierarchical Paradigm
2
•Chapter 3: Biological Foundations of the Reactive Paradigm
•Chapter 4: The Reactive Paradigm
•Chapter 5: Designing a Reactive Implementation
3
•Chapter 7: The Hybrid Deliberative/Reactive Paradigm
Team 1 Overview
Name:
Presents section of Book:
Jorge Franco
Introduction and Overview
Willmert Pereyra
What is a robot and brief
history 1.1 – 1.4.1
George Ragousis
Sylvester Delano
Robot Control and Operation
1.4.2 – 1.7
GPS Strips 2.1 – 2.2.3
Alexander Torres
NHC NIST RCS 2.2.4 – 2.7
Introduction and Overview
Jorge Franco
Overview
• What is AI robotics
– 3 major paradigms
• Ways in which intelligence is organized
• Architectures for paradigms
– Coherent
– Reusable
Implementations
• Single/Team of robots
What are Robots?
• Connotation/denotation
– anthropomorphic
• Origins on January 25, 1921, Prague, Karel
Capek’s play, R.U.R (Rossum’s Universal Robots)
• Term derived from Czech word “robota”, loosely
translated as menial worker.
– Attitude towards robot has disastrous consequences:
» Moral of rather socialist story: “Work defines a
person”
What are Robots? (cont’d)
• Shift from human-like servants made from
biological parts to human-like servants made up
of mechanical parts due to science fiction
– Classics:
• Metropolis (1926), The Day the Earth Stood Still (1951), and
Forbidden Planet (1956)
• Shift from human-like mechanical creatures to
whatever shape gets the job done is due to
reality
• Definition used in book: an intelligent robot is a
mechanical creature which can function
autonomously.
What are Robotic Paradigms?
• A paradigm is a philosophy or set of assumptions and/or
rules/techniques which characterize an approach to a class of
problems
• Why know paradigms?
– Key to successfully program a robot for an application
– Interesting from historical perspective
• Issues that spawned one the shift from one paradigm to another
• 3 kinds
– Hierarchical
– Reactive
– Hybrid deliberative/reactive
• Described in two ways
– Relationship between 3 accepted primitives
• Sense, Act, Plan
– Way that sensory data is processed and distributed through the system
Robot Paradigm Primitives
(fig1.2 from book)
Robot
Primitives
Sense
Plan
Act
Input
Output
Sensor data
Sensed information
Information (Sensed and/or
cognitive)
Directives
Sensed information/
directives
Actuator commands
Sensing Organization in Robot
Paradigms
• Way Sensory data:
– Processed
– Distributed
• Local processing
– Sensor information restricted to specific/dedicated
way for each robot function
• Global world model processing
– All SI first processed into a global world model
– Subsets of model distributed to other functions as
needed
Overview of the 3 Paradigms
fig.1.3 a.) Hierarchical, b.) Reactive, and c.) Hybrid
deliberative/reactive
a.
b.
c.
Hierarchical Paradigm
• 1967 – 1990
• Top down fashion – Heavy on planning
• Introspective view
– However as Cognitive Psych. now know:
• Not always good assessment of thought process.
• Default schemas or behaviors
• Global world model
– Hard and brittle
• Frame problem and closed world assumption
Another View of the Hierarchical
Paradigm (fig.1.4 from book)
Robot
Primitives
Sense
Plan
Act
Input
Output
Sensor data
Sensed information
Sensed and/or cognitive
information
Directives
Sensed information/
directives
Actuator commands
The Reactive Paradigm (fig.1.5 from book)
Robot
Primitives
Sense
Plan
Act
Input
Output
Sensor data
Sensed information
Sensed and/or cognitive
information
Directives
Sensed information/
directives
Actuator commands
The Hybrid Deliberative/Reactive
Paradigm (fig.1.6 from book)
Robot
Primitives
Plan
Sense-Act
(behaviors)
Input
Output
Information (Sensed
and/or cognitive)
Directives
Sensor data
Actuator commands
Representative Architectures
• Templates for an implementation
• Examples of what each paradigm really means
• According to Mataric: an architecture is a principled way
of organizing a control system, with constraints on the
way the control problem can be solved
• Common components in robot architecture and rules of
thumb for placing them together
– IC car –paradigm
– Each car manufacturer has its own architecture
– The car manufacturers may have slight modification on their
architecture for sedans, convertibles, SUV’s,etc.
Set Criteria for the Evaluation of an
Architecture
•
•
•
•
Modularity
Niche Targetability
Portability
Robustness
Layout of the Section
• Divided into 8 chapters
– 1. define Robotics
– 2. describes Hierarchical Paradigm and 2 architectures
– 3. sets the stage for understanding the Reactive Paradigm and
the motivation that spawned it.
– 4. Describes the Reactive Paradigm and popular architecture
– 5. Provides guidelines and case studies on designing robot
behaviors
– 6. Discusses simple sonar and computer vision processing
techniques
– 7. Describes the Hybrid Deliberative-Reactive Paradigm
– 8. Discusses how the principles of the 3 paradigms have been
transferred to team of robots
Sections 1.1 –1.4.1
Willmert Pereyra
Uses of Robots
• Dirty jobs.
• Dull jobs.
• Dangerous jobs.
Robotics Timeline
Planetary rovers
AI robotics
vision
Telemanipulators
1960
Telesystems
Industrial
manipulators
manufacturing
1970
1980
1990
2000
Old Movies About Robots
•
•
•
•
Modern Times (Charlie Chaplin), 1936.
Metropolis, 1927.
Silent Running, 1972.
The Phantom Menace, 1999.
Modern Times 1937
Metropolis 1927
Silent Running 1972
The Phantom Menace 1999
Approaches to Robotics
• Artificial Intelligence (AI).
• Engineering.
AI vs. Engineering
• AI:
– Uses paradigms.
– All actions are human-like.
• Engineering:
– Does not use paradigms.
– Actions performed are
mechanical.
Engineering Control Types
• Ballistic control:
– The position, trajectory and velocity profiles
are computed once.
• Feedback control:
– The error between the goal and current
position is noted by a sensor(s): a new
trajectory and profile is computed and
executed. Then modified in the next update.
AI Robotics Terms
• Intelligent Robot:
– A mechanical creature which can function
autonomously.
• Paradigm:
– A philosophy or set of assumptions and/or
techniques which characterize an approach to
a class of problems.
AI Robotics Terms
• Luddites:
– People who object to robots, or technology in
general.
• Artificial Intelligence (AI):
– (1) Science of making machines act
intelligently. (2) The study of ideas that enable
computers to be intelligent. (3) An attempt to
make computers do things that at present
people are better at.
AI Robotics Terms
• Teach pendant:
– A device that enables the programmer to
guide the robot through the desired set of
motions.
• Automatic Guided Vehicle (AGV):
– A vehicle that knows where it is, can plan a
path from its current location to its goal
destination and can avoid colliding with
obstacles.
AI Robotics Terms
• Telepresence:
– The reduction of cognitive fatigue and
simulator sickness by making the humanrobot interface more natural: virtual reality.
• Telemanipulator:
– Sophisticated mechanical linkage which
translates motions on one end of the
mechanism to motions at the other end.
AI Robotics Terms
• Industrial manipulator:
– A reprogrammable multifunctional mechanism
that is designed to move materials, parts,
tools, or specialized devices.
• Black factory:
– A factory that has no lights turned on because
there are no workers.
Architecture Evaluation Criteria
• Support for modularity:
– Good software engineering principles?
• Niche targetability:
– Works well for the intended application?
• Ease of portability:
– Works for other applications or other robots?
• Robustness:
– Is the system vulnerable? Where?
Model S Telemanipulator
Model S Telemanipulator
Movemaster Robot
Industrial Robots
Robotic Paradigms
1. Hierarchical.
2. Reactive.
3. Deliberative/Reactive.
Defining Paradigm Assumptions
• By the relationship between the primitives.
• By the way sensor data is processed and
distributed.
Global World Model Problems
• Constructing generic global world models
is very hard due to the frame problem and
the closed world assumption.
Global World Model Problems
• Frame problem:
– Deals with the representation of real-world
situations in a way that is computationally
tractable.
• Closed/Open world assumption:
– States that the world model contains
everything the robot needs to know (Closed)
and if it is violated the robot may not be able
to function correctly.
Hierarchical Paradigm
•
•
•
•
•
•
Oldest paradigm.
Prevalent from 1967-1990.
Robot operates top-down.
Emphasizes planning.
Assumes thought is introspective.
A global model captures all sensing
data.
Hierarchical Paradigm
Primitives
Input
Output
Sense
Sensor data
Sensed
information
Plan
Sensed and/or
Directives
cognitive information
Act
Sensed information/
directives
Actuator
commands
Hierarchical Paradigm
Robot Control and Operation
Section 1.4.2 – 1.7
George Ragousis
4 Ways to control and operate a robot
•
•
•
•
1. Remote control (RC)
2. Tele-operation
3. Semi-autonomous
4. Autonomous (AI)
1. Remote control
– you control the robot
– you can view the robot
and it’s relationship to
the environment
– operator isn’t
removed from scene,
not very safe
– ex. radio controlled
cars, bomb robots
Boxing RC robots 
2. Teleoperation
– you control the robot
– you can only view the
environment through
the robot’s eyes
– don’t have to figure
out AI
2. Teleoperation
Remote
Local
CommuniDisplay cation
Sensor
Mobility
Control
Effector
Power
Local
Remote
2. Teleoperation
•
•
•
•
•
•
is suitable for applications where…
the tasks are unstructured and not repetitive
the task workspace cannot be engineered to permit the
use of industrial manipulators
key portions of the task require dexterous manipulation,
especially hand-eye coordination, but not continuously
key portions of the task require object recognition or
situational awareness
the needs of the display technology do not exceed the
limitations of the communication link (bandwidth, time
delays)
the availability of trained personnel is not an issue
2. Teleoperation
•
•
•
•
Disadvantages…
Cognitive fatigue, 100% guidance
Simulator sickness
communications bandwidth (telepresence)
Time delays
(Darkstar 1 – Darkspot 0)
3. Semi-autonomous
Portion of directions and commands is given to robot
2 flavors
Shared control
step by step instructions
to accomplish task but no
full guidance is required
Control trading
commanding robot to do something
within its abilities and allowing the
robot to get it done without interaction
4. Autonomous
• Auto – nomous
auto = self
nomos = rule
self-commanded
• space robotics
• the need for autonomy
• artificial intelligence (AI)
Teleoperation Vs Autonomous
remote operation Vs self operation
• easy to achieve
• human in control – small
chances of decision and
judgment errors
• dexterous manipulations
• critical decisions by human
(Mars Pathfinder accident)
• Introduces time delays in
proportion with the distance
between local & remote.
• much more difficult to achieve
• higher risk of misjudgment and
false actions from robot
• no time delays in operation
• independent
goal of autonomy and AI:
To mimic the capabilities of animals or
humans sufficiently in order to survive
for long periods with only simple
instructions from earth.
Artificial Intelligence
1.
2.
3.
4.
5.
6.
7.
Seven areas
Knowledge representation – how am I me?
Understanding natural language (willing spirit – weak flesh)
Learning
Planning & problem solving
Inference – just take a decision
Search
Vision
Section 2.1 – 2.2.3
Sylvester Delano
The Hierarchical Paradigm
• Describe the Hierarchical Paradigm in terms of the
3 robot primitives and its organization of sensing
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
• Name and evaluate one representative
Hierarchical architecture in terms of: support for
modularity, niche targetability, ease of portability to
other domains, robustness
• Understand precondition, closed world
assumption, open world, frame problem
• List two advantages and disadvantages of the
Paradigm
Introduction to AIHierarchicalChapter2
Section 2.1 -2.2.3
Robotics (Team ONE)
58
Organization
SENSE
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
PLAN
ACT
World model:
1. A priori rep
2. Sensed info
3. Cognitive
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
59
Stanford Research Institute
●
●
SRI is an independent, non-profit research institute
conducting client-sponsored research and development for
government agencies, commercial businesses, foundations,
and other organizations.
SRI is well known for its innovations in communications and
networks, computing, economic development and science and
technology policy, education, energy and the environment,
engineering systems, pharmaceuticals and health sciences,
homeland security and national defence, and materials and
structures.
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
60
Shakey
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
• The first mobile
robot to be able to
reason about its own
actions, Shakey
combined research
in robotics, artificial
vision, and natural
language
processing.
• Built by SRI (Stanford
Research Institute) for
DARPA 1967-9
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
61
Shakey(cont'd)
• Programming was
primarily in
LISP.
• Used Strips as
main algorithm for
controlling what to
do
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
62
What is LISP(LIST
Processing) ?
• A high-level programming language used for
developing AI applications. Developed in 1960
by John McCarthy, its syntax and structure is
very different from traditional programming
languages. For example, there is no syntactic
difference between data and instructions.
• LISP is available in both interpreter and
compiler versions and can be modified and
expanded by the programmer. Many varieties
have been developed, including versions that
perform calculations efficiently.
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
63
Strips: Means-ends analysis
“Go to Stanford AI Lab”
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
INITIAL STATE:
Tampa, Florida (0,0)
GOAL STATE:
Stanford, California (1000,200)
Difference:
Introduction to AI
Robotics (Team ONE)
1020 miles
Chapter2 Section 2.1 -2.2.3
64
Difference Table
Distance
(difference)
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
mode of transportation
(OPERATOR)
d<=200 miles
FLY
100<d<200
TRAIN
d<=100
DRIVE
d<1
WALK
mode=difference_table(INITIAL STATE, GOAL STATE, difference)
Introduction to AI
Robotics (Team ONE)
1. Look up what to do: FLY
2. Not at SAIL, so repeat
3. Chapter2
LookSection
up what
to do: DRIVE
2.1 -2.2.3
65
Preconditions
difference
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
OPERATOR
PRECONDITIONS
d<=200 miles
FLY
100<d<200
TRAIN
d<=100
DRIVE (rental)
at airport
DRIVE (personal car)
at home
d<1
WALK
How do I know if I’m at the airport or at home?
Now must keep up with the state of the world
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
66
Maintaining State of the World:
Add and Delete Lists
distance
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
OPERATOR
PRECONDITIONS
d<=200
miles
FLY
100<d<2
00
TRAIN
d<=100
DRIVE
(rental)
at airport
DRIVE
(personal)
at home
d<1
Introduction to AI
Robotics (Team ONE)
ADD-LIST
DELETELIST
at city Y
at airport
at city X
at city Y
at train
station
at city X
WALK
Chapter2 Section 2.1 -2.2.3
67
Class Exercise
distance
Organization
-SPA
-global
Strips
-Shakey
Rep. Arch.
-evaluation
-NHC
-RCA
Summary
OPERATOR
PRECONDITIONS
d<=200
miles
FLY
100<d<2
00
TRAIN
d<=100
DRIVE
(rental)
at airport
DRIVE
(personal)
at home
d<1
Introduction to AI
Robotics (Team ONE)
ADD-LIST
DELETELIST
at city Y
at airport
at city X
at city Y
at train
station
at city X
WALK
Chapter2 Section 2.1 -2.2.3
68
Strips Summary
•
Designer must set up
– World model representation
Organization
-SPA
– Difference table with operators, preconditions, add & delete lists
-global
– Difference evaluator
Strips
• Strips assumes closed world
-Shakey
Rep. Arch.
– Closed world: world model contains everything needed for robot
-evaluation (implication is that it doesn’t change)
-NHC
– Open world: world is dynamic and world model may not be
-RCA
Summary complete
• Strips suffers from frame problem
– Frame problem: representation grows too large to reasonably
operate over
Introduction to AI
Robotics (Team ONE)
Chapter2 Section 2.1 -2.2.3
69
Section 2.2.4 –2.7
Alexander Torres
Team One – Hierarchy
STRIPS Summary
•
Designer must set up
– World model representation
– Difference table with operators, preconditions, add &
delete lists
– Difference evaluator
•
Strips assumes closed world
– Closed world: world model contains everything needed
for robot (implication is that it doesn’t change)
– Open world: world is dynamic and world model may not
be complete
•
Strips suffers from frame problem
– Frame problem: representation grows too large to
reasonably operate over
Team One – Hierarchy
Closed World Assumption and the Frame Problem
It is impractical for a programmer to come up with all possible
reactions, conditions to all probable cases in the real world
The need to formally represent the world and then maintain every
change about it is nonnutritive.
The axioms (facts) that would frame the world would quickly
become too numerous for any realistic domain
A proposed solution was ABStrips which divided the problem into
multiple layers of abstraction (this would mean solving problems
with increasing levels of details)
Team One – Hierarchy
Nested Hierarchical Controller (NHC)
Representative Architecture
Nested Hierarchical Controller (NHC)
•SENSE
•PLAN
•ACT
The robot gathers observation from its
sensors and combines that information
with priori knowledge to create the World
Model.
From the World Model, the robot can
PLAN what action it should take.
Team One – Hierarchy
Nested Hierarchical Controller (NHC)
Representative Architecture
Planning for navigation consists of three
step executed by
Mission Planner, Navigator, and Pilot
Each of these can access the World
Model
The last step is the Pilot module
generating specific actions for the robot to
do.
Team One – Hierarchy
Nested Hierarchical Controller (NHC)
The Benefits of the NHC are:
•Unlike STRIPS it interleaves
planning and acting
•It can adapt to changes in its
environment if necessary
The Disadvantages of NHC are:
•Planning Function is only appropriate
for navigation tasks
Team One – Hierarchy
NIST REAL Time Control System RCS
Real-time Control System Architecture
Created by Jim Albus
Best suited for semi-autonomous control
Based on NHC, RCS is developed as a
guide for manufacturers who wish to add
AI to their robots.
Sensory perception modules introduce a
useful preprocessing step between the
sensor and the fusion into a world model
The Value Judgment module simulates
the plan to ensure they work.
Behavior Generation Module operates
similar to the pilot with less focus on
navigations.
Team One – Hierarchy
Advantages and Disadvantages
Advantage
• Provides an ordering of the relationship between sensing, planning, and
acting.
Disadvantages
• Planning, every update cycle the robot would have to update a global world
model and do some type of planning.
• Sensing and action are disconnected. This doesn’t allow for reflexive
reactions found in real life.
• Dependence on global world model is related to the frame problem. A
simple task can becomes incredibly complicated to describe.
• Uncertainty in semantics, sensor noise and actuator errors.
Team One – Hierarchy
Programming Considerations
•
Predicate logic and recursion used by STRIPS favors languages such as
LISP and PROLOG
•
Although LISP and PROLOG do not have good real-time control properties,
the alternative at the time was FORTRAN IV which did not support
recursion
•
Hierarchical Paradigm forces programming for specific tasks instead of
object oriented tasks.
•
NHC and RCS decomposition of a task is not modular in design
Team One – Hierarchy
Summary
•Except for NIST Real-time Control Architecture, Hierarchical Paradigm has
fallen out of favor for more biologically based systems of control.
•It has contributed concepts and terminology such as preconditions,
closed/open world assumptions, and the frame problem
•It has the inherent property to allow an evolution of intelligence from semiautonomous control to full autonomy.