Transcript ppt - LaDiSpe - Politecnico di Torino

ROBOTICS
01PEEQW
Basilio Bona
DAUIN – Politecnico di Torino
Mobile & Service Robotics
Supervision and control
Traditional approach
Traditional artificial intelligence considers robot “brains” as
sequential processing units
Main ideas
 Representation -> reasoning -> planning
 Model building (maps of the environment is required)
 Functional decomposition; hierarchical systems
 Symbolic/semantic manipulation
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
3
Supervision and Control
Perception
Localization &
Map Building
Data
treatment
Task/mission commands
Position
Global map
iteration
Path planning
& Reasoning
Data
treatment
commands
data
Sensors
Actuators
Motion control
a priori knowledge
Environment
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
4
Supervision and Control
Localization
Map Building
Position
Global map
Local map
World model
Path planning
Reasoning
Path
Motion
control
Perception
Environment
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
5
Control Strategies
 Control loop requirements
 World changes dynamically
 A compact model of the world is very difficult to define
 There are many sources of uncertainty, both in the world and in the
robot
 Two possible approaches + a combination of them
 Classic AI – deliberative model




Approximate world modeling (model-based method)
Based on a set of functions
Vertical decomposition
Top-down approach
 Modern AI – reactive model
 No world model is given: behavior-based control
 Horizontal decomposition
 Bottom-up approach
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
6
Control Characteristics
Sense – Plan – Act
Subsumption/Reactive model
This architecture may produce a
slow and delayed response
Function 1
use model
plan
act
Basilio Bona - DAUIN - PoliTo
Horizontalapproach
Vertical approach
sense
Function 2
Function 3
Function 4
Function 5
ROBOTICS 01PEEQW - 2015/2016
7
Model-Based Approach
 Requires the complete modeling of the world
 Each block is a computed function
 Vertical decomposition and nested-embodiment of functions
An example
sensors
Perception
Localization - Map building
Cognitive planning
Motion control
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
actuators
8
Model-Based Approach
Planner
GOAL RECOGNITION
GLOBAL PATH PLANNING
Navigator
SUB-GOAL FORMULATION
LOCAL PATH PLANNING
Pilot
TARGET GENERATOR
DYNAMIC PATH PLANNING
Path monitor
TARGET LOCATION
PATH CORRECTION/OBSTACLE AVOIDANCE
Controller
COMMANDS
Low level control
SENSORS
ACTUATORS
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
9
Behavior-Based Approach
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
10
Behavior-Based Approach
Rodney Brooks, inventor of this approach, writes:
 Complex behavior needs not necessarily be the product of a complex control
system
 The world is its best model
 Simplicity is a virtue
 Intelligence is in the eye of the observer
 Robots should be cheap
 Robustness in the presence of noisy or failing sensors is a design goal
 Planning is just a way of avoiding figuring out what to do next
 All onboard computation is important
 Systems should be built incrementally
 No representation. No calibration. No complex computers. No high band
communication
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
11
Behavior-Based Approach
Various names given to this approach
 Subsumption architecture
 Reactive system
 Reflexive behavior
 Perception-action
Action 2
Perception 2
Perception 1
Action 1
WORLD
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
12
Subsumption architecture
 The subsumption architecture was originally proposed by Brooks
[1986]
 The subsumption architecture copies the synergy between
sensation and actuation in lower animals such as insects
 Brooks argues that instead of building complex agents in simple
worlds, we should follow the evolutionary path and start building
simple agents in the real, complex and unpredictable world
 From this argument, a number of key features of subsumption
result
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
13
Subsumption architecture
1. No explicit knowledge representation is used. Brooks often refers to
this as “The world is its own best model”
2. Behavior is distributed rather than centralized
3. Response to stimuli is reflexive – the perception-action sequence is
not modulated by cognitive deliberation
4. The agents are organized in a bottom-up fashion. Thus, complex
behaviors are fashioned from the combination of simpler,
underlying ones
5. Individual agents are inexpensive, allowing a domain to be
populated by many simple agents rather than a few complex ones.
These simple agents individually consume little resources (such as
power) and are expendable, making the investment in each agent
minimal
6. Several extensions have been proposed to pure reactive
subsumption systems. These extensions are known as behaviorbased architectures
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
14
Behavior-Based Approach
Characteristics
 No model is necessary
 Horizontal decomposition
 Coordination + Priority = Fusion
 Biomimesis = observe and copy animal behavior
 Subsumption
 Embodiment
Avoid obstacles
Discover new areas
sensors
Detect goal position
Communicate data
å
actuators
Recharge
Follow right/left wall
Coordination / fusion
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
15
Control Strategies: deliberative vs reactive
DELIBERATIVE
Model-based
REACTIVE
Behavior-based
Purely symbolic
Reflexive
Speed of response
Predictive capabilities
Depends on accurate world models
• Depends on the world
representation
• Slow response
• High level intelligence (cognition)
• Variable latency
Basilio Bona - DAUIN - PoliTo
• Representation-free
• Real-time response
• Low level intelligence (stimulusresponse)
• Fast and easy computation
ROBOTICS 01PEEQW - 2015/2016
22
Embodiment
 To embody (verb) = to manifest or personify in concrete form; to
incarnate; to incorporate, to unite into one body
 Embodiment is the way in which human (or any other animal)
psychology arises from the brain & body physiology
 Embodiment theory was introduced into AI by Rodney Brooks in the
‘80s. Brooks have claimed that all autonomous agents need to be
both embodied and situated
 The theory states that intelligent behavior emerges from the
interplay between brain, body and world. Brain, body and world are
equally important factors in the explanation of how particular
intelligent behaviors originate in practice
 Brooks showed that robots could be more effective if they “thought”
(planned or processed) as little as possible
 The robot's intelligence is organized only for handling the minimal
amount of information necessary to make its behavior be
appropriate and/or as desired by its creator
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
23
Embodiment
 Rolf Pfeifer (AILab Zurich) says that there are essentially two
directions in artificial intelligence:
 one focused on developing useful algorithms or robots;
 and another direction that focuses on understanding intelligence,
biological or artificial
 In order to make progress on the second direction, an embodied
perspective is mandatory
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
24
Situated robot
 A situated robot is one which does not deal with abstract
representations of the world (which may be simulated or real),
but rather reacts directly to its environment as seen through its
sensors
 An alternative to having a situated robot would be one which
builds up a representation of its world and then makes plans
based on this representation
 Because of the limitations of our present technology, these two
approaches often seem contradictory
 At present, each approach can be appropriate for different
applications
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
25
Situated robot
 The situated approach is good for dealing with problems where
planning ahead is unnecessary or takes too much time
 However, the representation approach is needed for solving more
complicated problems where it is necessary to reason about the
state of the world
 For dealing with complicated tasks in the real world, it will probably
be necessary to fuse the two approaches
 Reasoning can be used to build up higher level plans and solve high
level problems
 Lower level functions may use a more situated approach for carrying
out plans and dealing with problems which need immediate
attention
 The structure and relations that originates from the interaction of
simple controllers and complex environment is called emergent
behavior
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
26
Robotics and AI
Main areas of Artificial Intelligence applied to robotics
1.
2.
3.
4.
5.
6.
7.
Knowledge representation
Understanding natural languages
Learning
Planning and problem solving
Inference
Search
Vision
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
27
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
28
Knowledge representation
 Define, build and memorize the physical and virtual structures
used by the robot to represent
 the world
 the desired tasks
 itself
 Example: a robot is looking after a human being under the
wreckage of a fallen building: how it is represented?
 Structural model:
 Head (oval) + trunk (cylindrical) + arms (cylindrical)
 Bilateral symmetry
 Physical model (thermal image, …)
 What happens if the body is only partially visible?
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
29
Understanding natural languages
Natural language is one of the most simple and human ways to interact
But …
To understand the words does not mean to understand the meaning
Grammatical structure vs Semantic structure
Example
We gave the monkeys two bananas because they were hungry
We gave the monkeys two bananas because they were over-ripe
They have the same grammatical structure, but a very different
semantic structure
To understand the sense we must know both the monkeys and the
bananas
Necessity to develop ontologies
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
30
Ontology
 An ontology is a formal representation of knowledge as a set of
concepts within a domain, and the relationships between those
concepts. It is used to reason about the entities within that domain,
and may be used to describe the domain
 An ontology is a “formal, explicit specification of a shared
conceptualization.” An ontology provides a shared vocabulary, which
can be used to model a domain — that is, the type of objects and/or
concepts that exist, and their properties and relations
 Ontologies are the structural frameworks for organizing information
and are used in artificial intelligence, etc., as a form of knowledge
representation about the world or some part of it
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
31
Learning
 Learning is the capacity to memorize actions and behaviors and to
repeat them to adapt to the implicit or explicit objectives
 In a broad sense, learning is the ability to adapt during life
 We know that most living organisms with a nervous system display
some type of adaptation during life
 The ability to adapt quickly is crucial for autonomous robots that
operate in dynamic and partially unpredictable environments, but the
learning systems developed so far have so many constraints that are
hardly applicable to robots that interact with an environment without
human intervention
 Learning requires




A structure able to store and retrieve data
One or more explicit objectives
An adaptation mechanism (reward + punishment)
An explicit or implicit teacher
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
32
Planning and problem solving
 Intelligence is associated to the ability to plan actions toward the
the given task fulfillment, and to solve problems arising when
plans fail
Go there
Basilio Bona - DAUIN - PoliTo
Go there
ROBOTICS 01PEEQW - 2015/2016
33
Planning and problem solving
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
34
SLAM
SLAM
Simultaneous Localization and Mapping
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
35
Inference
 Inference is a procedure that allows to generate an answer also
when the available data or information are incomplete
 Inference is based on statistical models (bayesian networks) or
semantic models
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
36
Search
 Search does not necessarily mean a true search of objects in
space, but defines the ability to examine a knowledge
representation data-base (search space) to find the required
answer
 Consider a computer playing chess: the best move is found
looking for a solution in the search space of all possible moves,
starting from the present chessboard state
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
37
Vision
 Vision is the most important sense in human beings
 Psychological studies have demonstrated that the ability to
solve problems is due to our brain capacity to visualize the
effects of each action
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
38
Books
R.C. Arkin
Behavior-Based Robotics
MIT Press, 1998
R.R. Murphy
Introduction to AI
Robotics
MIT Press, 2000
Basilio Bona - DAUIN - PoliTo
G. Dudek, M. Jenkin
Computational Principles of
Mobile Robotics
Cambridge U.P., 2000
R. Siegwart, I.R. Nourbakhsh
Autonomous Mobile Robots
MIT Press, 2004
ROBOTICS 01PEEQW - 2015/2016
39
Books
Rolf Pfeifer, Josh C. Bongard
How the Body Shapes the Way We
Think A New View of Intelligence
Foreword by Rodney Brooks
MIT Press, 2006
S. Thrun, W. Burgard, D. Fox
Probabilistic Robotics
MIT Press, 2005
Autori Vari
Principles of Robot Motion
MIT Press, 2005
Basilio Bona - DAUIN - PoliTo
ROBOTICS 01PEEQW - 2015/2016
40