MOBI LE & SERVI CE ROBOTI CS ROBOTI CS CFI DV CA 01 - - PowerPoint PPT Presentation

CY 02CFI C

MOBI LE & SERVI CE ROBOTI CS

CFI DV

ROBOTI CS

CA – 01

Supervision and control

OBOTI C

Basilio Bona

RO

DAUIN – Politecnico di Torino

Basilio Bona – DAUI N – Politecnico di Torino 001/ 1

Supervision and Control

CY

L li ti Path planning Position Global map a priori knowledge Task/mission commands

02CFI C

Localization Map Building Path planning reasoning p

CFI DV

Data Data

• l

CA – 01

Data treatment Data treatment data rception

• n contro

commands

OBOTI C

Sensors Actuators data Per Motio

RO

Environment

Basilio Bona – DAUI N – Politecnico di Torino 001/ 2

Environment

Supervision and Control

CY

Position

02CFI C

Localization Map Building Path planning Reasoning

Global map

CFI DV

Local map World model Path

CA – 01

Perception Motion control

OBOTI C

control Environment

RO

Environment

Basilio Bona – DAUI N – Politecnico di Torino 001/ 3

Control Strategies St t f th t l l

CY

Structure of the control loop

World changes dynamically A compact model of the world does not exist 02CFI C A compact model of the world does not exist There are many sources of uncertainty, both in the world and in the robot

T ibl h

CFI DV

Two possible approaches

– Classic AI – deliberative model Complete modeling (model based method) CA – 01 Complete modeling (model-based method) Function based Horizontal decomposition OBOTI C Top-down approach – Modern AI – reactive model No world model: behavior-based RO No world model: behavior-based Vertical decomposition Bottom-up approach

Basilio Bona – DAUI N – Politecnico di Torino 001/ 4

Control Strategies

CY DELI BERATI VE Model-based REACTI VE Behavior-based

Purely symbolic Reflexive

02CFI C

Speed of response

Purely symbolic Reflexive

CFI DV

Predictive capabilities

CA – 01

Depends on accurate world models

OBOTI C

• Depends on the world representation
• Slow response

High level intelligence (cognition)

• Representation-free
• Real-time response

Low level intelligence (stimulus response)

RO

• High level intelligence (cognition)
• Variable latency
• Low level intelligence (stimulus-response)
• Fast and easy computation

Basilio Bona – DAUI N – Politecnico di Torino 001/ 5

Control Characteristics

Sense Plan Act Subsumption/Reactive model

CY

Sense – Plan – Act This architecture may prevent a fast and timely response Subsumption/Reactive model

http://ai.eecs.umich.edu/cogarch0/subsump/

02CFI C

and timely response

CFI DV

sense ch Task 1

CA – 01

use model approac Task 2 Task 3

OBOTI C

plan Vertical a Task 3 Task 4

RO

act V Task 5

Basilio Bona – DAUI N – Politecnico di Torino 001/ 6

Model-Based Approach

CY

• Complete modeling of the world
• Each block is a computed function

02CFI C

• Vertical decomposition and nested-embodiment of functions

CFI DV

An example Perception

sensors

CA – 01

Localization - Map building

OBOTI C

Cognitive planning

RO

Motion control

actuators

Basilio Bona – DAUI N – Politecnico di Torino 001/ 7

Model-Based Approach

A d l t d b di t

CY

A second example: nested embodiment

High level mission

02CFI C

High level mission Service Task

CFI DV

Servo Elemental move Motion primitives

CA – 01

Servo

OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 8

Model-Based Approach

A third example: nested embodiment

CY

Planner

GOAL RECOGNITION GLOBAL PATH PLANNING

02CFI C

Navigator

SUB-GOAL FORMULATION LOCAL PATH PLANNING

CFI DV

Pilot

TARGET GENERATOR DYNAMIC PATH PLANNING

CA – 01

Path monitor

TARGET LOCATION PATH CORRECTION/OBSTACLE AVOIDANCE

OBOTI C

Low level control Controller

COMMANDS

RO

Low level control

SENSORS ACTUATORS

Basilio Bona – DAUI N – Politecnico di Torino 001/ 9

Behavior-Based Approach R ti t

CY

Reactive systems Reflexive behavior Perception-action

02CFI C

Perception action Subsumption

CFI DV

ROBOT

CA – 01

ROBOT

Perception 1 Perception 2 Action 1 Action 2

OBOTI C

WORLD

p

RO

WORLD

Basilio Bona – DAUI N – Politecnico di Torino 001/ 10

Behavior-Based Approach

Rodney Brooks is the father of this approach: CY Rodney Brooks is the father of this approach:

Some of his key sentences

02CFI C

• Complex behavior need not necessarily be the product of a complex

control system ll h f h b CFI DV

• Intelligence is in the eye of the observer
• The world is its best model
• Simplicity is a virtue

CA – 01

• Simplicity is a virtue
• Robots should be cheap
• Robustness in the presence of noisy or failing sensors is a design goal

OBOTI C

• Planning is just a way of avoiding figuring out what to do next
• All onboard computation is important

S t h ld b b ilt i t ll RO

• Systems should be built incrementally
• No representation. No calibration. No complex computers. No high

band communication

Basilio Bona – DAUI N – Politecnico di Torino 001/ 11

Behavior-Based Approach

CY

No model is necessary Horizontal decomposition C di ti P i it F i

02CFI C

Coordination + Priority = Fusion Biomimesis = observe and copy animal behavior Subsumption

CFI DV

Subsumption Embodiment

CA – 01 OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 12

Subsum ption The subsumption architecture was originally

CY

The subsumption architecture was originally proposed by Brooks [ 1986] .

02CFI C

The subsumption (or 'Brooksian') architecture is based on the synergy between sensation and actuation in lower animals such as insects

CFI DV

actuation in lower animals such as insects. Brooks argues that instead of building complex

CA – 01

g g p agents in simple worlds, we should follow the evolutionary path and start building simple agents in h l l d di bl ld

OBOTI C

the real, complex and unpredictable world. From this argument, a number of key features of

RO

From this argument, a number of key features of subsumption result:

Basilio Bona – DAUI N – Politecnico di Torino 001/ 13

Subsum ption

1 No explicit knowledge representation is used Brooks

CY

• 1. No explicit knowledge representation is used. Brooks
• ften refers to this as “The world is its own best model”

2 h d b d h h l d

02CFI C

• 2. Behavior is distributed rather than centralized.
• 3. Response to stimuli is reflexive – the perception-action

CFI DV

p p p sequence is not modulated by cognitive deliberation 4 The agents are organized in a bottom up fashion Thus

CA – 01

• 4. The agents are organized in a bottom-up fashion. Thus,

complex behaviors are fashioned from the combination of simpler, underlying ones

OBOTI C

• 5. Individual agents are inexpensive, allowing a domain to

be populated by many simple agents rather than a few

RO

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, ki th i t t i h t i i l

Basilio Bona – DAUI N – Politecnico di Torino 001/ 14

making the investment in each agent minimal

Subsum ption

CY

Several extensions (Mataric, 1992) have been proposed to pure reactive subsumption systems.

02CFI C

These extensions are known as behavior-based architectures.

CFI DV

architectures. Capabilities of behavior-based systems include landmark detection and map building learning to

CA – 01

landmark detection and map building, learning to walk, collective behaviors with homogeneous agents, group learning with homogeneous agents, and

OBOTI C

g p g g g , heterogeneous agents.

RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 15

Em bodim ent T b d ( b) if t if i t

CY

To embody (verb) = manifest or personify in concrete form; incarnate; incorporate, unite into one body Em bodim ent is the way in which human (or any other

02CFI C

Em bodim ent is the way in which human (or any other animal) psychology arises from the brain & body physiology

CFI DV

physiology It is specifically concerned with the way the adaptive function of categorization works, and how things acquire

CA – 01

g , g q names It is distinguished from developmental psychology and

OBOTI C

physical anthropology by its focus on cognitive science,

• ntogeny, ontogenetics, chaos theory and cognitive

notions of entropy far more abstract and more reliant

RO

notions of entropy – far more abstract and more reliant

• n mathematics

Basilio Bona – DAUI N – Politecnico di Torino 001/ 16

Em bodim ent

CY

Embodiment theory was brought into AI by Rodney Brooks in the 1980s B k d th h d th t b t ld b

02CFI C

Brooks and others showed that robots could be more effective if they “thought” (planned or processed) and perceived as little as possible

CFI DV

and perceived as little as possible The robot's intelligence is geared towards only handling the minimal amount of information

CA – 01

handling the minimal amount of information necessary to make its behavior be appropriate and/ or as desired by its creator

OBOTI C

Brooks (and others) have claimed that all autonomous agents need to be both embodied and it t d Th l i th t thi i th l t

RO

• situated. They claim that this is the only way to

achieve strong AI

Basilio Bona – DAUI N – Politecnico di Torino 001/ 17

Em bodim ent (R lf Pf if AIL b Z i h) th ti ll t

CY

(Rolf Pfeifer AILab Zurich) there are essentially two directions in artificial intelligence: one concerned with developing useful algorithms or robots; and another

02CFI C

developing useful algorithms or robots; and another direction that focuses on understanding intelligence, biological or otherwise.

CFI DV

In order to make progress on the latter, an embodied perspective is mandatory In this research branch

CA – 01

perspective is mandatory. In this research branch, artificial intelligence is embodied.

OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 18

Som e term s O t i th t b h f lif i hi h d l ith

CY

Ontogeny is that branch of life science which deals with the study of origin and development of an organism from fertilized ovum to its mature form In more general

02CFI C

fertilized ovum to its mature form. In more general terms, ontogeny is defined as the history of structural change in a unity, which can be a cell, an organism, or a

CFI DV

society of organisms, without the loss of the organization that allows that unity to exist

CA – 01

Em bodied Em bedded Cognition, a position in cognitive science stating that intelligent behavior emerges out of

OBOTI C

the interplay between brain, body and world Em bodied Cognition (or the embodied mind thesis), a

RO

Em bodied Cognition (or the embodied mind thesis), a position in cognitive science and the philosophy of mind emphasizing the role that the body plays in shaping the mind

Basilio Bona – DAUI N – Politecnico di Torino AA 2007/ 08 001/ 19

Situated robot

CY

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

02CFI C

real), but rather reacts directly to its environment as seen through its sensors. An alternative to having a situated robot would be one

CFI DV

An alternative to having a situated robot would be one which builds up a representation of its world and then makes plans based on the representation.

CA – 01

Because of the limitations of our present technology, these two approaches often seem contradictory.

OBOTI C

In the present, each approach is better for different applications.

RO

If we want to make an “artificial person” at some point in the future, we will need to incorporate both approaches.

Basilio Bona – DAUI N – Politecnico di Torino 001/ 20

Situated robot

CY

The situated approach can only deal with a small domain

• f problems.

02CFI C

When a robot gets into a situation where it needs to reason or plan ahead in order to reach a goal, simply ti t it i t i i ffi i t

CFI DV

reacting to its environment is insufficient. A situated robot can be thought of as sensors and goals

CA – 01

g g feeding into a fixed network of difference engines which have actuators as their outputs.

OBOTI C

For example, we might have a light sensors and a goal

• f reaching lights fed into a difference engine.

RO

The robot would then activate its actuators in an attempt to reach its goal.

Basilio Bona – DAUI N – Politecnico di Torino 001/ 21

Situated robot

CY

This network might be slightly more complex with a controller suppressing and activating difference engines in

02CFI C

co t o e supp ess g a d act at g d e e ce e g es response to an FSM state or a sensor input (for example, causing the robot to flee when danger is detected).

CFI DV

However, this architecture does not give us the flexibility we need to solve more complicated problems, such as figuring out that we need to move further from some

CA – 01

figuring out that we need to move further from some goals in order to reach an overall goal. A it t d b t i ht h i l b ilt i t it

OBOTI C

A situated robot might have special cases built into it (e.g. for dealing with getting stuck in a corner or down a dead-end hallway), but it would be very difficult to deal

RO

y), y with the general case this way.

Basilio Bona – DAUI N – Politecnico di Torino 001/ 22

Situated robot

CY

The situated approach is good for dealing with problems where planning ahead is unnecessary or takes too much time

02CFI C

time. However, the representation approach is needed for solving more complicated problems where it is necessary

CFI DV

solving more complicated problems where it is necessary to reason about the state of the world. F d li ith li t d t k i th l ld it

CA – 01

For dealing with complicated tasks in the real world, it will probably be necessary to fuse the two approaches.

OBOTI C

Reasoning can be used to build up higher level plans and solve high level problems while lower level agencies may use a more situated approach for carrying out plans and

RO

use a more situated approach for carrying out plans and dealing with problems which need immediate attention.

Basilio Bona – DAUI N – Politecnico di Torino 001/ 23

Robotics and AI Th t t d l ti th t i i t f th

CY

The structure and relations that originates from the interaction od simple controllers and complex environment is called emergent behavior

02CFI C

environment is called emergent behavior Seven areas of AI applied to robotics

– Knowledge representation CFI DV Knowledge representation – Understanding natural languages – Learning CA – 01 – Planning and problem solving – Inference – Search OBOTI C Search – Vision RO

R.R. Murphy, Introduction to AI Robotics, MIT Press, 2000.

Basilio Bona – DAUI N – Politecnico di Torino 001/ 24

Know ledge representation D fi d b ild th h i l d i t l t t d

CY

Define and build the physical and virtual structures used by the robot to represent

– the world 02CFI C the world – the tasks – itself

E l b t i l ki ft h b i d

CFI DV

Example: a robot is looking after a human being under the wreckage of a fallen building: how it is represented?

– Structural model: CA – 01 – Structural model: Head (oval) + trunk (cylindrical) + arms (cylindrical) Bilateral symmetry OBOTI C – Physical model (thermical image?) – What happens if the body is only partially visible? RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 25

Understanding Natural Languages

CY

Natural language is one of the most simple and human ways to interact But

02CFI C

But … Understand the words does not mean to understand the sentence

Grammatical structure vs Semantical structure CFI DV Grammatical structure vs Semantical structure

Example

CA – 01

We gave the monkeys two bananas because they were hungry We gave the monkeys two bananas because they were over-ripe

OBOTI C

They have the same grammatical structure bu a very differente semantical structure To understand the sense we must know both the monkeys and the bananas

RO

To understand the sense we must know both the monkeys and the bananas Necessity to develop ontologies

Basilio Bona – DAUI N – Politecnico di Torino 001/ 26

F l th t i Ti fli lik Ti fli lik b i t t d i

Understanding Natural Languages

CY

For example, the string Time flies like an arrow Time flies like an arrow may be interpreted in a variety of ways:

02CFI C time moves quickly just like an arrow does; measure the speed of flying insects like you would measure that

• f an arrow - i.e., (You should) time (verb) flies like you would

CFI DV

, ( ) ( ) y an arrow;

measure the speed of flying insects like an arrow would - i.e.

Time flies in the same way that an arrow would (time them).;

CA – 01

Time flies in the same way that an arrow would (time them).;

measure the speed of flying insects that are like arrows - i.e.

Time those flies that are like arrows;

a type of flying insect "time-flies " enjoy arrows (compare Fruit OBOTI C a type of flying insect, time-flies, enjoy arrows (compare Fruit

flies like a banana.) The word “time” alone can be interpreted as three different parts of

RO

The word “time” alone can be interpreted as three different parts of speech, (noun in the first example, verb in 2, 3, 4, and adjective in 5). English is particularly challenging in this regard because it has little

Basilio Bona – DAUI N – Politecnico di Torino AA 2007/ 08 001/ 27

g p y g g g inflectional morphology to distinguish between parts of speech.

Learning

CY

• Is the capacity to memorize actions and behaviors and to repeat

them to adapt to the implicit or explicit objectives I b d l i i h bili d d i lif 02CFI C

• 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 CFI DV display some type of adaptation during life.

• The ability to adapt quickly is crucial for autonomous robots

that operate in dynamic and partially unpredictable CA – 01 p y p y p 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 OBOTI C interact with an environment without human intervention.

• Learning requires

– A structure able to store and retrieve data

RO

– 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 – DAUI N – Politecnico di Torino 001/ 28

An explicit or implicit teacher

Planning and problem solving

CY

Intelligence is associated to the ability to plan actions toward the the given task fulfillment and to solve

02CFI C

toward the the given task fulfillment, and to solve problems arising when plans fail

CFI DV CA – 01 OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 29

Go there Go there

Planning and problem solving

CY 02CFI C CFI DV CA – 01 OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 30

SLAM

CY 02CFI C CFI DV CA – 01 OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 31

I nference

CY 02CFI C

Inference is a procedure that allows to generate an answer also when the data or information are

CFI DV

incomplete Inference is based on statistical models (bayesian

CA – 01

Inference is based on statistical models (bayesian networks) or semantical models

OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 32

Search

CY

Not necessarily a search of objects in space, but the

02CFI C

ability to examine a knowledge representation data- base (search space) to find the required answer

CFI DV

Consider a computer playing chess: the best move is found looking for a solution in the search space of

CA – 01

g p all possible moves, starting from the present chessboard state

OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 33

Vision

CY 02CFI C

Vision is the most important sense in human beings

CFI DV

Psychological studies have demonstrated that the ability to solve problems is due to our brain

CA – 01

y p capacity to visualize the effects of each action

OBOTI C RO

Basilio Bona – DAUI N – Politecnico di Torino 001/ 34

Ontology

In philosophy ontology listed as a part of the major branch of CY

• In philosophy, ontology listed as a part of the major branch of

philosophy known as metaphysics, ontology deals with questions concerning what entities exist or can be said to exist, and how such b d l d h h h d bd d d 02CFI C entities can be grouped, related within a hierarchy, and subdivided according to similarities and differences CFI DV

• It derives from Greek οντος, “òntos” (present participle of ειναι,

“einai”, to be) and λογος, "lògos". CA – 01

• In computer sciences, an ontology is a formal representation of a

set of concepts within a domain and the relationships between those concepts It is used to reason about the properties of that domain OBOTI C

• concepts. It is used to reason about the properties of that domain,

and may be used to define the domain.

• An ontology provides a shared vocabulary which can be used to

RO

• 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.[

Basilio Bona – DAUI N – Politecnico di Torino 001/ 35

Books

CY 02CFI C

R.C. Arkin

CFI DV

Behavior-Based Robotics MIT Press, 1998

CA – 01 OBOTI C

R R M h

RO

R.R. Murphy Introduction to AI Robotics MIT Press, 2000

Basilio Bona – DAUI N – Politecnico di Torino 001/ 36

Books

CY 02CFI C

• G. Dudek, M. Jenkin

Computational Principles of Mobile

CFI DV

Robotics Cambridge U.P ., 2000

CA – 01 OBOTI C

• R. Siegwart, I.R. Nourbakhsh

RO

Autonomous Mobile Robots MIT Press, 2004

Basilio Bona – DAUI N – Politecnico di Torino 001/ 37

Books

CY 02CFI C

Autori Vari

CFI DV

Autori Vari Principles of Robot Motion MIT Press, 2005

CA – 01 OBOTI C

How the Body Shapes the Way We Think

A N Vi f I t lli RO A New View of I ntelligence

Rolf Pfeifer and Josh C. Bongard Foreword by Rodney Brooks

Basilio Bona – DAUI N – Politecnico di Torino 001/ 38

Books

CY 02CFI C

• S. Thrun, W. Burgard, D. Fox

CFI DV

Probabilistic Robotics MIT Press, 2005

CA – 01 OBOTI C

Rolf Pfeifer, Josh C. Bongard How the Body Shapes the Way We Think A N Vi f I t lli

RO

A New View of Intelligence Foreword by Rodney Brooks MIT Press, 2006

Basilio Bona – DAUI N – Politecnico di Torino 001/ 39