Emergence and Development of Embodied Cognition

EDEC-2001
Workshop at the International Conference on Cognitive Science ICCS2001

The objective of this workshop is to bring together researchers from cognitive science,
psychology, robotics, artificial intelligence, philosophy and related fields to discuss the
role of developmental and embodied views of cognition, and in particular, their mutual
relationship. The ultimate goal of this approach is to understand the emergence of high-
level cognition in organisms based on their interactions with the environment over
extended periods of time.

Program
13:30 – 13:50 Intro Rolf Pfeifer
13:50 – 14:25 Talk 1 Kazuo Hiraki
14:25 – 15:00 Talk 2 Max Lungarella
15:00 – 15:20 Break
15:20 – 15:55 Talk 4 Giorgio Metta
15:55 – 16:30 Talk 5 Stefan Schaal
16:30 – 17:30 Discussion and poster session with drinks to ensure a relaxed
atmosphere

Talks
Kazuo Hiraki (Department of Systems Science, University of Tokyo, Japan): Causality
and Prediction: Detection of Delayed Intermodal Contingency in Infancy

Max Lungarella (Artificial Intelligence Lab, University of Zurich, Switzerland): Robots

as Cognitive Tools: An Information-theoretic Analysis of Sensor-(motor) Data

Giorgio Metta (Humanoid Robotics Group, MIT-Artificial Intelligence Lab, Cambridge,

MA, USA): Development in Artificial Systems

Stefan Schaal (Department of Computer Science, University of Southern California, LA,

CA, USA and Computational Learning Group, ATR Labs, Kyoto, Japan): Movement
Primitives and Imitation Learning Based on Dynamic System Theory
Posters
Charles Lenay, Olivier Gapenne, and John Stewart: The constitution of spatiality in
relation to the lived body: a study based on prosthetic perception

Hideki Kozima and Hiroyuki Yano: In search of ontogenetic prerequisites for

embodied social intelligence

Andres Perez-Uribe and Michele Courant: A robotics framework for the study of
signal coevolution

Catholijn M. Jonker, Jacky L. Snoep, Jan Treur, Hans V. Westerhoff, and Wouter
C.A. Wijngaards: Embodied intentional dynamics of bacterial behavior

Nick Barnes: Three embodied vision-guided docking methods for mobile robots

Max Lungarella and Rolf Pfeifer: Robots as cognitive tools: an information-theoretic

analysis of sensory-motor data

Program Committee
Rolf Pfeifer (chair, AI Lab, University of Zurich, Switzerland)
Max Lungarella (AI Lab, University of Zurich, Switzerland)
Yasuo Kuniyoshi (Department of Mechano-Informatics, University of Tokyo, Japan)
Olaf Sporns (Department of Psychology, Indiana University, Bloomington, IN, USA)
Giorgio Metta (Humanoid Robotics Group, MIT AILab, Cambridge, MA, USA)
Giulio Sandini (LIRA-Lab, University of Genova, Italy)
Rafael Nunez (University of California, Berkeley, USA)
 Causality and Prediction: Detection of Delayed Intermodal
Contingency in infancy
Kazuo Hiraki1,2, Naoko Dan 2 and Sotaro Shimada2
1
Department of Systems Science, University of Tokyo <khiraki@idea.c.u-tokyo.ac.jp >
2
Presto, JST <{dan, shimada}@ardbeg.c.u-tokyo.ac.jp>
Abstract
The paper presents our recent project to study
infant’s detection of proprioceptive-visual
contingency. As a key idea for the study, we
introduced temporal delay between actions and
consequent stimulation. Six -month-old infants were
presented with either an on-line (no-delay) view or a
delayed view of their own legs on a large screen. The
preliminary result using preferential looking method
suggests that six-month-old infants can discriminate
between no- delay and delayed views of their leg
motion even the delay time is quite short (2 sec).
Development of contingency detection
How do infants distinguish themselves from external
world? Detection of intermodal contingency plays an
important role to distinguish the sensory consequences of
self-produced actions from externally produced sensory
stimuli. The contingency detection can be seen as a
fundamental ability for the early development of self.
Also it can be a base for much higher-level cognitive
process such as prediction or perception of causality.
Recent evidence in developmental cognitive
psychology revealed that even young infants have
so-called ecological self. For example, Bahrick and
Watson(1985) used preferential looking method to study
self-exploration for a part of infants body (i.e. the legs).
They demonstrated that 5-month-olds are sensitive to
proprioceptive-visual contingency. In their experiment,
the infant was placed in front of two TV monitors, side by
side. On one monitor, the infants saw a contingent view
of their own legs. The other monitor showed either a
noncontingent, prerecorded tape of the baby's own legs
or a tape of another baby's leg movements. Bahrick and
Watson(1985) showed that 5-month-olds look
preferentially to the noncontingent view.
Rochat and Morgan(1995) conducted the similar
experiments but they systematically manipulated spatial
relationships; the temporal contingency between the 2
images and the infant's actual movements was
maintained constant. In their studies, 3- and
4-5-month-old infants looked longer at different
perspectives of their legs (left-right inversion) and
generated more leg activity while looking at those view s.
The results of these studies can be interpreted as
evidence of early detection of intermodal invariants.
However, a question still remains as to infant ’s temporal
sensitivity, which is crucial for defining what the
contingency is. Because there exist a lots of delayed or
lagged stimuli that are consequence of actions, we have
to take into account the delayed contingency for the
further understanding of the infant’s cognition. We can
recognize the response of a computer as a consequence of
our action (i.e. clicking a mouse button) even if it is very
slow and lagged responses are employed. Bahrick and
Watson(1985) used the prerecorded image, but it had
almost no relationship between infant’s action. In order
to study the temporal sensitivity we have to use much
shorter delay and systematically manipulate it.
Delayed contingency: preliminary result
As a key idea to study the infant ’s detection of
proprioceptive-visual contingency, we introduced
temporal delay between actions and consequent visual
stimulation to extend the previous studies.
Six-month-old infants were presented with two views on
a large screen. One was the contingent on-line (no-delay)
view as used in the previous studies. The other view was
also contingent but delayed two seconds from the on -line
view. A special devise was used to generate the delay on
the image from a CCD camera capturing infant’s leg
motion.
Figure 1 summarizes the looking time data for every
one-minute period during the three minutes session and
the total time. As the figure indicates, six-month-old
infants looked longer at the delayed view for each period.
The result suggests that they can discriminate between
the no-delay and the delayed views of their leg motion,
even if the delay time is quite short (2 sec).
 plausibility. However, questions remain as to when the
45 discrepancy between actual and predicted sensory stimuli
is compared inside the brain, and how the internal model
40
develops. For these questions, we believe that the
35 developmental study for infant’s temporal sensitively to
30 contingency provides key ideas.
25
20 on-line
delay Re ferences
15
10
Bahrick, L. E. and Watson, J.S.(1985): Detection of
5 intermodal proprioceptive-visual contingency as a
0 potential basis of self-perception in infancy.
1 3 Developmental Psychology, 21-6, 963--973.
min. min.

Figure 1. Looking time for on-line and Blakemore S-J., Frith C.D. and Wolpert, D.M.(1999):
Spatio-temporal prediction modulates the perception of
delayed view. self -produced stimuli. Journal of Cognitive Neuroscience,
11:5, 551-559.

It is worth while to note that the delayed view shown in Rochat,P. and Morgan,R.(1995): Spatial Dterminants in
our experiment were far from the noncontingent view the Perception of Self-Produced Leg Movements by 3- to
shown in Bahrick and Watson(1985) . However, it is 5-Month-Old Infants. Developmental Psychology. Vol.
difficult to conclude that infants in our study 31, No.4, 626-636.
distinguished the delayed contingency from the
noncontingency because the infants in Bahrick and Watson,J.S.(1993): Detection of self: The perfect
Watson(1985) also exhibited preference to the algorithm. In S. T. Parker, R. W. Mithcell, and M. L.
noncontingent views. In order to investigate whether or Boccia (Eds.), Self-awareness in animals and humans:
not infants can discriminate delayed view from Developmental perspectives, 131-148.
noncontingent view, we are currently conducting
additional experiments by using either a pair of delayed
and noncontingent view or a pair of on-line and longer
delayed view. Through these studies, we may get much
more convinced implication.

Towards a computational model of

prediction
Although the result reported above is preliminary and
more work needs to be done, we believe that our attempt
offers a new idea to construct a computational model of
prediction and/or perception of causality during infancy.
Watson(1993) proposes four theoretical
options--contiguity, correlation, conditional probability,
and causal implication--as the focal relation between an
infant’s behavior and consequent stimulation. Although
the options are suggestive, it is still unclear at the
computational level; how infant’s brain discriminate the
sensory consequences of self-produced actions from
externally produced sensory stimuli.

Blakemore, Frith and Wolpert(1999) proposed a model

for determining the sensory consequences of a movement
in the study to investigate why self -produced tactile
stimulation is perceived as less intense than the same
stimulus produced externally. Their prop osed model is an
internal forward model that makes predictions of the
sensory feedback based on the motor commands to the
actuator (i.e. efference copy). The forward models are
attractive because they provide the computational
Emergence and Development of Embodied Cognition (EDEC2001)
CONCEPTUAL INTEGRATION
Gilles Fauconnier
Department of Cognitive Science, UCSD
Cognitive science research in the last
twenty-five years has provided
considerable evidence that reason is
embodied. The neural architectures that
evolved to produce perception,
sensation, and bodily movement are at
the heart of what we experience as
rational inference, conceptualization,
and meaning construction. Metaphor
theory and mental space theory played a
significant role in showing the
inadequacies of the abstract, algorithmic,
and disembodied views that dominated
structuralism, generative linguistics, and
logic-based truth-conditional approaches
to semantics. Conceptual integration
theory, often called "blending" or CI, is
a further development of this line of
research. It confirms in novel ways that
similar general properties of neural
binding and simulation lie behind
sensorimotor activities, concrete
interaction with the world, human-scale
everyday experience, abstract reasoning,
and scientific or artistic invention.
CI is based on extensive empirical
observation in multiple areas of meaning
construction. Detailed proposals have
been made regarding its constitutive and
governing principles, and on the
remarkable compressions and emergent
dynamics they allow. A substantial
body of research now exists on CI in
mathematics [9, 13], social science [14,
16], literature [5, 15], linguistics [10,
11], and music [18]. There have been
proposals for the mathematical and
computational modeling of the operation
[6, 17], and experimental research has
been carried out on the corresponding
neural and cognitive processes [1, 7].
Additional information and an extensive
bibliography can be found on the
website <blending.stanford.edu>.
The findings have raised
fundamental research questions of
interest to all of us, and in particular to
the participants of the present workshop.
What follows is a general overview of
the issues, results, and research program.
It deliberately avoids going into
technical detail. Each of the mentioned
phenomena has received extensive
analysis elsewhere.
CI is a basic mental capacity that
leads to new meaning, global insight,
and conceptual compressions useful for
memory and manipulation of otherwise
diffuse ranges of meaning. It plays a
fundamental role in the construction of
meaning in everyday life, in the arts and
sciences, in technological development,
and in religious thinking. The essence of
the operation is to construct a partial
match between input mental spaces and
to project selectively from those inputs
into a novel 'blended' mental space,
which then dynamically develops
emergent structure. It has been
suggested that the capacity for complex
 2

conceptual blending ("double-scope" than a clipper that sailed the same course
integration) is the crucial capacity in 1853. A sailing magazine reports:
needed for thought and language.
As we went to press, Rich Wilson and Bill
Most of our thinking, even in the
Biewenga were barely maintaining a 4.5 day lead
simplest circumstances, is unbelievably over the ghost of the clipper Northern Light,
complex but mercifully completely whose record run from San Francisco to Boston
unconscious. Our conscious experience they're trying to beat. In 1853, the clipper made
is one of direct simplicity, in the same the passage in 76 days, 8 hours. —"Great
America II," Latitude 38, volume 190, April
way that our conscious perception of
1993, page 100.
objects ("a blue cup") is one of direct
simplicity that bypasses (in
There are two distinct events in this
consciousness) what the neural circuits
story, the run by the clipper in 1853 and
of the brain are painstakingly integrating
the run by the catamaran in 1993 on
behind the scenes. CI is no exception to
(approximately) the same course. In the
the laws of backstage cognition: it too
magazine quote, the two runs are merged
operates largely behind the scenes,
into a single event, a race between the
choreographing vast networks of mental
catamaran and the clipper's "ghost". The
spaces beyond the reach of our
two distinct events correspond to two
conscious awareness, yielding cognitive
input mental spaces, which reflect salient
products at the conscious level which
aspects of each event: the voyage, the
appear straightforward and
departure and arrival points, the period
unproblematic.
and time of travel, the boat, its positions
Here are some well-known, highly
at various times. The two events share a
visible examples of conceptual blends
more schematic frame of sailing from
repeated here for illustrative purposes:
San Francisco to Boston; this is a
Boat race
"generic" space, which connects them.
A famous example of blending is
Blending consists in partially matching
"the boat race" or "regatta". A modern
the two inputs and projecting selectively
catamaran is sailing from San Francisco
from these two input spaces into a fourth
to Boston in 1993, trying to go faster
mental space, the blended space:

Generic space
cross-space mapping
Input space 1
selective projection
Blended space
In the blended space, we have two
boats on the same course, that left the
starting point, San Francisco, on the
same day. Pattern completion allows us
to construe this situation as a race (by
importing the familiar background frame
of racing and the emotions that go with
it). This construal is emergent in the
blend. The motion of the boats is
structurally constrained by the
mappings. Language signals the blend
explicitly in this case by using the
expression "ghost-ship." By "running
the blend" imaginatively and
dynamically — by unfolding the race
through time — we have the relative
positions of the boats and their
dynamics.
Here is another attested example
where embodied conceptual blending
leads to emergent bodily movement. A
ski instructor was trying to teach a
novice to hold his arms correctly and
look down the slope (rather than in the
direction of the skis). The instructor told
3
Input space 2
the novice to imagine that he was a
waiter in Paris carrying a tray with
champagne and croissants. By focusing
his attention on the tray, and trying to
avoid spilling the champagne, the novice
was able to produce something
approaching the right integrated motion
on the slope. The Inputs in this case are
the ski situation and the restaurant
situation, with arm and body positions
mapped onto each other. The Generic
space has only human posture and
motion, with no particular context. But
in the Blend, the skier is also actually
carrying the champagne tray. The
blended space is in a way a fantasy, but
it allows a real motion to emerge.
Rather remarkably, this pedagogical feat
requires no actual champagne and
croissants. Just thinking of them is
enough.
Conceptual blending with emergent
structure shows up in all areas of human
behavior; cultures develop successive
blends and the ones that become

entrenched can be transmitted to new
generations. This is the case in the
evolution of mathematical concepts and
the development of mathematics itself.
For example, whole numbers are
blended with one dimensional space, the
output is blended with proportions,
yielding the rationals. Blending with
geometric constructs leads to the
irrationals, and then to transcendent
numbers. Blending with two-
dimensional space yields negative and
then complex numbers. At every step,
there is rich emergent structure which
mathematicians explore. The discovery
of non-euclidean geometries follows
similar paths. These processes are
analyzed in great detail in [3], [4], [9],
[13].
The recent technology of the mouse
and desktop computer interface is
another case in which a succession of
blends are integrated conceptually and
physiologically into a single activity
with considerable emergent structure. A
CI network based on perceptual
invariance, coherence, stability, and
non-ubiquity integrates two dimensional
illuminations on the screen with 3D
objects. This is the object blend.
Another CI network integrates the 3D
horizontal and sometimes jerky
manipulation of the mouse with the
perceived motion of the arrow (an
"object" by virtue of the object blend) on
the screen in a vertical plane. The
embodied capacity to integrate the
manipulation of the mouse with the
motion of the arrow - to feel that one is
4
moving the arrow - is remarkable. It is
quickly acquired by simple linking and
coactivation of the input mental spaces
and their frames. Other blends for the
interface develop in the same way: they
include a grasping and moving blend
and a containment blend. The
successive blends and their emergent
structue make it possible to experience
the clicking on the rectangle as a form of
grasping, and the motion of the clicked
mouse as deliberately moving an object.
This looks easy when we do it but
demands highly complex correlations
and integrations at many levels: neural,
perceptual, motor, conceptual.
COMPRESSIONS AND VITAL
RELATIONS
Compression in blending networks
operates on a set of relations rooted in
fundamental human neurobiology as it
applies to shared physical and socio-
cultural human experience. These vital
relations, which include Cause-Effect,
Change, Time, Identity, Intentionality,
Representation, and Part-Whole, can
apply across mental spaces, and they
also define essential topology within
mental spaces. In blending networks, a
vital relation across inputs (outer-space
vital relation) can be compressed into a
vital relation within the blended space
(inner-space vital relation). There are
systematic compression hierarchies, such
as: ANALOGY, DISANALOGY > CHANGE,
IDENTITY > UNIQUENESS. Blending, it
turns out, is an instrument of

compression par excellence. Some
spectacular cases of compression and
decompression of rich vital relation
structures are analyzed in [4].
GRADIENT OF COMPLEXITY
A more detailed examination of CI
networks reveals a continuum of
complexity with four prototypes that
stand out on that continuum: Simplex,
Mirror, Single-Scope, and Double-
Scope. It turns out that ordinary
compositional logic corresponds to the
straightforward blending found in
Simplex networks. Anything beyond
that will be truth-functionally non-
compositional. The point is crucial,
since compositionality is often taken to
be a necessary prerequisite to the
analysis of meaning. At the high end of
the continuum of blending complexity,
we find Double-Scope networks which
blend inputs with different (and often
clashing) organizing frames to produce
creative emergent frame structure in a
blended space. Double-scope blending
is what we typically find in scientific,
artistic, and literary discoveries and
inventions. Double-scope creativity is
perhaps the most striking characteristic
of our species.
MATERIAL ANCHORS
Some of the things that we often take
to be the most basic in everyday living
and thinking are the result of creative
successive blends evolved by cultures
over time. Edwin Hutchins [8] has
shown that material objects are
5
used/designed by cultures as material
anchors for conceptual blends.
Manipulation of these objects is a way of
running complex blends. Intricate
successive blending leads to watches,
coins or bills, and rational numbers. As
the anchors change so do our concepts of
time, money, and number. Watches are
material anchors (in Hutchins' sense) for
a powerful conceptual blend that
compresses an outer-space linear
ordering of successive days that can go
on to infinity into an inner-space cyclical
ordering of repeated motion through the
same unique day. This integration
network provides one of the inputs (the
Cyclical Blended Day) to another
integration network, the Timepiece
network, from which our modern
everyday conception of time emerges.
Money is a material anchor that also
reflects the construction of successive
intricate blends. The basic notion of
exchange leads to an implicit and
abstract value scale. Blending the input
of Values with an input of Goods
augmented by material objects like bills
and coins yields a blended space of
Goods and Money with Values in which
the elementary structure of buying and
selling as we know it can emerge. And
further successive blends will lead to
ever more complex emergent financial
and economic dynamic organizations. A
major goal of a social science like
economics is to understand the emergent
structures that result.

LEARNING
Learning by new generations is
possible because much of the
manipulation of the blended space can
be learned without having to consciously
apprehend the full networks of which it
is part. Although buying and selling, or
telling the time are activities that take
time for children to learn, it is not the
case that the children need to go through
the long explorations and the many
garden paths that the culture went
through before it came up with watches
or money.
During the first three years of the
child's life, a seemingly impossibly
complex system of elaborate blending
networks is constructed. The child has a
biological capacity for double-scope
integration and the culture has a range of
specific integration networks to offer.
The biology and the culture combine to
impressive effect. Once a fundamental
culturally-motivated integration network
is in place, its compressed, human scale
blend seems as obvious and inevitable as
the perception of a blue cup.
Communicating with language, making
and recognizing representations, and
using a fork and a spoon seem
straightforward both for the child and the
adults who interact with the child. The
fact that combinations of vocal sounds
mean what they mean, that the marks on
the paper are a dog, and that the fork is
an instrument for eating are taken as
direct and inevitable. We can no longer
6
see the mark without seeing the dog, or
the fork without seeing its purpose and
its etiquette. This is exceptionally
difficult: a human brain in its first few
years is the quickest and most plastic
and most capacious complex system in
the universe, but it takes at least three
years of constant working to bring it to
mastery of all these culturally-motivated
blends. This exceptionally difficult
development is almost entirely
unconscious for the child and mostly
unobservable by adults, who see only
superficial signs of the mental activity of
the child and take them for granted
because adults do not remember the
difficulty. Just as biology gives us
entrenched integrations that we can
manipulate directly, so culture and
learning result in the building of
entrenched integrations that we can
manipulate directly. In both cases, once
we have the integration, it is hard or
impossible to escape it. We construe the
physical, mental, and social worlds we
live in by virtue of the integrations we
achieve through biology and culture.
There is no other way for us to
apprehend the world. Conceptual
integration is not something we do in
addition to living in the world. Instead,
it is one of the essential means we have
for apprehending and constructing our
world. Human beings have reached a
double-scope level of conceptual
integration and this accounts for many of
their specific performances.
 7

REFERENCES

[1] Coulson, Seana. 2001. Semantic Leaps: Frame-shifting and Conceptual Blending in Meaning
Construction. New York and Cambridge: Cambridge University Press.
[2] Fauconnier, Gilles. 1997. Mappings in Thought and Language. New York and Cambridge:
Cambridge University Press.
[3] Fauconnier, Gilles and Turner, Mark . 1998. "Conceptual Integration Networks." Cognitive Science.
Volume 22, number 2 (April-June 1998), pages 133-187.
[4] Fauconnier, Gilles and Turner, Mark. in press. The Way We Think. New York: Basic Books.
[5] Freeman, Margaret. 1997. "Grounded spaces: Deictic-self anaphors in the poetry of Emily Dickinson,"
Language and Literature, 6:1, 7-28.
[6] Goguen, Joseph. 1999. "An Introduction to Algebraic Semiotics, with Application to User Interface
Design." Computation for Metaphor, Analogy, and Agents. Edited by Chrystopher Nehaniv. Berlin:
Springer-Verlag, pages 242-291. A volume in the series Lecture Notes in Artificial Intelligence.
[7] Grush, Rick and Nili Mandelblit. 1997. "Blending in language, conceptual structure, and the cerebral
cortex." The Roman Jakobson Centennial Symposium: International Journal of Linguistics Acta Linguistica
Hafniensia Volume 29:221-237. Per Aage Brandt, Frans Gregersen, Frederik Stjernfelt, and Martin Skov,
editors. C.A. Reitzel: Copenhagen.
[8] Hutchins, E. in preparation. "Material Anchors for Conceptual Blends"
[9] Lakoff, G. and Nuñez, R. 2000. Where Mathematics Comes From. New York: Basic Books.
[10] Liddell, Scott K. 1998. "Grounded blends, gestures, and conceptual shifts." Cognitive Linguistics, 9.
[11] Mandelblit, Nili. 1997. Grammatical Blending: Creative and Schematic Aspects in Sentence
Processing and Translation. Ph.D. dissertation, UC San Diego.
[12] Oakley, Todd. 1998. "Conceptual blending, narrative discourse, and rhetoric." Cognitive Linguistics,
9-: 321-360.
[13] Robert, Adrian . 1998. "Blending in the interpretation of mathematical proofs." Discourse and
Cognition.. Edited by Jean-Pierre Koenig. Stanford: Center for the Study of Language and Information
(CSLI).
[14] Sweetser, Eve. 2000. "Blended Spaces and Performativity". Cognitive Linguistics. Vol. 11, 3/4.
[15] Turner, Mark. 1996. The Literary Mind. New York: Oxford University Press
[16] Turner, Mark. 2001. Cognitive Dimensions of Social Science. New York: Oxford University Press.
[17] Veale, Tony. 1999. "Pragmatic Forces in Metaphor Use: The Mechanics of Blend Recruitment in
Visual Metaphors." Computation for Metaphor, Analogy, and Agents. Edited by Chrystopher Nehaniv.
Berlin: Springer-Verlag, pages 37-51. A volume in the series Lecture Notes in Artificial Intelligence.
[18] Zbikowski, Lawrence. in press. The Conceptual Structure of Music. New York: Oxford University
Press.
[19] Websites for Mental Spaces and Conceptual Blending:
http://www.mentalspace.net http://www.blending.stanford.edu
 Development in artificial systems
Giorgio Metta (pasa@ai.mit.edu)
AI-Lab – Massachusetts Institute of Technology, Cambridge MA, USA

Giulio Sandini (giulio@dist.unige.it)

LIRA-Lab – DIST, University of Genova, Genova, Italy

Lorenzo Natale (nat@lira.dist.unige.it)

LIRA-Lab – DIST, University of Genova, Genova, Italy

Riccardo Manzotti (manzotti@lira.dist.unige.it)

LIRA-Lab – DIST, University of Genova, Genova, Italy

Francesco Panerai (francesco.panerai@college-de-france.fr)

LPPA-Collége de France, Paris, France

Abstract skill. Although, this has been successful in some cases

AI and robotics tackled the problem of building autonomous
creatures (human-like sometimes) from many different
directions. In spite of this, our systems are still primitive, little
adaptive and prone to failure. We present a possible
alternative to the construction of embodied artificial systems
where the process of development is explicitly considered.
This is in contrast with the common practice of modular
design. We describe also an instantiation of this principle in
an anthropomorphic robot interacting and acting in a real
environment. The robot is equipped with diverse sensory
modalities such as vision, proprioception, audition, and it
interacts with the environment through a sophisticated motor
apparatus.
The approach
Sensori-motor development in biological systems is a
structured process, where all the sensing mechanisms
are gradually integrated with the actuating mechanisms
in order to achieve a mature performing system. An
extremely interesting aspect of this process is its
underlying ability to deal with and solve complex
multi-level integration problems. From an engineering
point of view, development can be thought of as a
hierarchy of integration processes which involve at
different times different action systems. The whole
process leads each single newborn through intermediate
developmental stages characterized by different level of
functionalities and capabilities. Eventually the
procedure results in the achievement of the final
performing and mature individual (Piaget, 1936; Thelen
& Smith, 1998).
In robotics, the attempts to build complex robots (e.g.
humanoids robots) have always followed principles of
modular design: often, to make the problem tractable,
engineers applied a “divide and conquer” strategy.
Following this approach, each module deals with a
specific sensory modality and/or a particular motor
(Brooks, 1996; Hirai, Hirose, Haikawa, & Takenaka,
1998), the final integration proved to be very difficult,
and often failed. Moreover, the system was assumed to
be time-invariant and consequently adaptation and
learning were not explicitly taken into account.
This paper presents an alternative design
methodology, which although in its infancy, it is suited
to deal with the complexity, adaptation, and integration
issues of artificial and biological systems. Taking into
account the lesson learned from the modular design
approach and trying to overcome the problem of the
final integration, we propose the use of engineered
principles of development to build complex systems.
The aim of the theory should be, firstly, to achieve
efficient and adaptive robots and, secondly, to develop
a “neurobotic toolbox” in which ideas about neural
computation can be tested by engineering working
systems. The latter issue should be considered of
particular interest for both robotics and neuroscience
due to the possibility to foster an interdisciplinary
approach to the study of complex systems.
The theory should address two level of analysis: i)
system analysis proper, where the system is described,
constituents individuated and the functioning explained,
and ii) construction theory: that is, how to design a
working system; how to take into account time-
variance, adaptation and learning. The first issue has
been mainly the approach taken by biologists; the latter
concerned mostly engineers – we propose a tighter
integration of the two fields.
The study of the biology – the modeling of brain
functions – can certainly suggest how to build more
successful and adaptable “artificial beings”. Adaptation
raises the issue of learning; in other words, how can the
learner acquire useful information in order to
accomplish a given task? Which sensors does it need?
Is learning always feasible? Until now, robotics and AI
have failed to give a definitive answer (assuming it
does exist) and indirectly they have also failed to
produce truly autonomous and flexible agents. In spite
of many successes in building robots of various shape,
size, abilities, sensory types, etc. there seems to be
something lacking in terms of “cognitive abilities”, as
well as adaptability of the system to the dynamic of the
environment.
In recent times “brain sciences” also face an
increasingly and intricate picture, where it is hard to
discover the general underlying principles, which
eventually will bear the real explanatory power. The
panorama consists of a huge number of brain areas and
intricate interconnections between them (the so-called
“telephone switchboard” model). Here lies the
significance of modeling: many researchers have
employed computational models to explain functions,
to derive general rules, and to integrate data gathered by
using different methodologies. Perhaps, not everything
is suitable to be simulated, and the world – the
“external” environment – is something, which is too
complex to be replicated appropriately. In this sense a
Figure 1: Three shots of the experimental setup. Left: complete view of the robot; middle: close up view of the head,
note the cameras and microphones with earlobes; right: rear view of the robot and in particular of the gyros.
Outlines of a theory
The first hints about what to include in a theory of
developing systems comes from statistics and learning
theory. The main result we refer to is derived from
Statistical Learning Theory (Vapnik, 1998). Vapnik et
al. suggest that the key of learning in a realistic setting
is complexity control. That is, not only the parameters
of the learning machine need to be optimized, but the
structure of the machine as well. We suggest that
development is about controlling complexity. This
approach justifies also the empirical observation that an
increase of complexity is observed during development
(Turkewitz & Kenny, 1982). The presence of regulatory
robot can be a suitable tool where to condense our
knowledge of brain functions and to test our models,
ultimately, against a real environment.
In this context, development can provide a vantage
point, for observing the temporal evolution of particular
skills, in particular, to determine, what prerequisites,
drives, and teaching signals are needed to learn them.
This offers the clearest view on learning, not only, in
algorithmic terms, but also about its logical
consistency.
The following section outlines some of the
foundations of a possible theory. Although some of the
elements are at the moment speculative, testing and
verification are part of the ongoing research.
Further, we describe an instantiation of the approach
in a behaving developing agent, which learns sensori-
motor coordination during the interaction with the
environment. We are aware that this is not yet complete
but by adopting a learning by doing philosophy, we
hope, in the future, to frame the approach in a more
consistent mathematical framework.
processes in the brain, which change the complexity of
the learner, have been documented (Johnson, 1997)
although rarely interpreted in this way (Quartz &
Sejnowski, 1997).
To put development in an “ecological” context, we
need further to consider how training data is collected.
This is because, data does not come for free: a cost
might be associated to the effort of collecting it. Action
(and embodiment) needs to be explicitly considered.
The recent analysis of the premotor cortices and the
discovery of, for example, mirror neurons (Di
Pellegrino, Fadiga, Fogassi, Gallese, & Rizzolatti,
1992; Rizzolatti & Fadiga, 1998) cast a new light in the
necessity of action as a prior to cognitive functions.
Another element of the theory is thus the view of action
as a foundation. Aspects of the development of the
ventral vs. dorsal stream also support this notion.
Kovacs (Kovacs, 2000), in fact, pointed out that the
control of movements (dorsal stream) develops before
the ability to categorize (ventral stream). Other
examples can be found in (Atkinson, 1998).
Although not immediately obvious, the theory
requires, at an even more fundamental level, goal-
directness and causal understanding. In fact, the very
basic element of every learning procedure is that of
detecting regularities in the sensory data together with
the ability to link the effects with the appropriate
causes. Goal-directness has to be seen as a way to
understand when the result of an action is within a
particular class irrespective of the viewpoint or sensory
modality – classes are, for example, grasping, holding,
tearing, etc. Any low-level sensory-motor coordination
task (e.g. reaching, eye movements, etc.) can be learnt
in principle by repetitive trials. The efferent copy and
the sensory reafferences generated during action need
to be appropriately linked and synchronized if the cause
and effect ought to be correctly interpreted. Appropriate
mapping can fill in the gaps (i.e. the transformations)
and their inverses to recover the correct movement to
apply in a given situation. This procedure can only take
into account the learning of approximately direct
mappings between the sensory and the motor space.
The cause-effect principle has conversely a much
broader applicability.
At the moment, we regard these considerations,
mostly as speculative, because a real theory should be
backed up by appropriate data. To recap, the goals of
defining a theoretical developmental framework are:
i) Engineered principled: define what are the
necessary elements of the theory of learning.
Examples we put forward are complexity
control, the primacy of action, and
understanding of causality. The latter has to
be intended in practical terms rather than in
general ones.
ii) Definition of a neurobotic toolbox: provide
an environment where to test biologically
plausible theories of development. It is not
required to be humanoid-shaped, although
we believe it has to be an embodied device
with the ability to act in the real world. We
already discussed, why we need action as a
constituent of the theory.
iii) Ongoing investigation is devoted mostly to
the identification of the constituents of the
model. From the robotic side, a model
incorporating some of the discussed learning
and development aspects has been realized
(see next section).
iv) It is clear that more experiments need to be
carried out in order to answer specific
questions developmentally. This might
include, in the future, experiments with
animals or human subjects. A measure of the
success of the method is certainly the
possibility of suggesting particular aspects
of the theory that need to be worked out, or
conversely, where the biological
explanations are missing. The farther we
will get along this line, the more successful
the integrated approach we are proposing.
Delivering the promises (or part of them)
The artificial system, named Babybot (Metta, Sandini,
& Konczak, 1999; Sandini, Metta, & Konczak, 1997),
consists of a twelve degree-of-freedom eye-head-arm
platform equipped with vision, audition, proprioception
and vestibular senses. The robot acquires and processes
images in a space variant format also known as log-
polar (Sandini & Tagliasco, 1980). The robot’s eyes
observe the world through a high-resolution fovea and a
lower resolution periphery. Acoustic sensation is
provided by a couple of microphones and plastic
earlobes. The vestibular sensor consists of three gyros
arranged along mutually orthogonal axes and it
provides a measure of the angular velocity of the head
in space (Panerai, Metta, & Sandini, 2000). A global
view of the setup is shown in figure 1, together with a
close-up view of the head and the vestibular sensor.
The process we focused on is how Babybot learns to
coordinate the eyes, head and arm from an initial stage
where control is mostly reflex-based. Figure 2 shows
the temporal dependences of the different maps
controlling the Babybot and where learning is applied.
The specific learning tasks the robot is involved in
(see (Metta, 2000)) steer the system from the initial
reflex-driven stage, with the help of some noise (Metta,
Carlevarino, Martinotti, & Sandini, 2000), toward more
stable patterns of sensori-motor coordination. For
example, initially the robot learns the relationship
between eye movements and image displacements,
which is needed to correctly foveate a spotted location,
but also to move the system into a second stage. In fact,
from the initial short eye movements the robot learns
how to generate correct saccades. As the eyes get under
control, more complicated coordination patterns might
emerge, the neck can be moved, and thus head-eyes
coordination can be acquired. This latter stage is
accomplished by learning both the proper eye to head
coordination map, but also a vestibulo-ocular reflex
(VOR). VOR itself is fine tuned automatically on the
basis of a visuo-motor performance measure based on
optical flow. Intuitively if the whole coordination
schema is not performing well, the lack of stability is
translated into a residual optical flow, which drives a
corresponding teaching signal tailored at reducing the
instability. In parallel, the system practice also with
reaching by trying roughly to touch whatever object is
located in its field of view. Of course, at least initially,
most of these tentative movements are wrong. Babybot,
though, improves by measuring its errors (only visual in
this case), and after some hundreds of trials learns how
to reach appropriately within the arm’s workspace. It is
worth noting that vision is not the only available
sensory modality; sound for example is used to redirect
the attention of the robot toward possibly interesting
events. Also in this case, maps linking the spatial
perception of sound with action systems are learnt.
Further, audition and vision are integrated in a coherent
percept, which allows the robot to consistently redirect
gaze when both sound and vision are present or when
only one of them is available.
As mentioned before, every sensory modality
available to the robot is activated from the very
beginning. Conversely the sequence of possible action
systems is limited by various factors (i.e. only the eyes
are appropriately coordinated at the very beginning, the
neck moves subsequently, and the torso later in time
(Bertenthal & Von Hofsten, 1998)). This dependence
shapes in a particular way both learning and the kind of
data gathered while learning.
The different “cortical” maps of Babybot are tuned in
order to work together and integration is already
embedded into the system. The sequence of learning
events is constrained in part by the physics of the
system, in part by how the modules are designed. The
resulting maps can develop in many different ways
depending on the particular history of learning itself.
For example, a lack of vestibular stimulation, or a
failure of the sensor itself, would lead to a reduced
coordination of the eye-head system. As a consequence,
the gain of the movement would be automatically
retuned because of the diminished stability and
increased residual optic flow. This is a very diverse
adaptation with respect to a system where the different
parts are designed in isolation. Further, because the
maps are embodied and mutually interacting the state
space experienced by the learner might rest in a sub-
manifold of the actual parameter space: i.e. not all
combinations of parameters are valid or physically
generable. The benefit is that the exploration space
might be reduced as well as the time required to acquire
particular skills.
Conclusion and future work
In conclusion, development might be a necessary
characteristic of adaptive systems: in other words, the
particular sequence of developmental events simplifies
learning by constraining the possible “roads” towards
skilled behavior. For example, newborns do not control
all available degrees of freedom but rather start by
practicing only a few. External/physical (like gravity)
and internal/neural limitations (like reflexes or initial
synergies) effectively shape how learning has to be
carried out.
From the theoretical point of view, statistical learning
theory (Vapnik, 1998) actually endorses a procedure
where the complexity of the system is controlled or
increased as a function of the “experience” of the
system (i.e. the number of training samples). A
developing system might take advantage of this
situation by increasing the complexity of the controller
over time. This insight can be also employed at a
different level of analysis to justify a growth of
complexity of the controlling neural networks (Quartz
& Sejnowski, 1997). This is not in contrast with
prevailing theories on neural selection (Edelman, 1988).
The key aspect in fact is the control of complexity, thus,
two processes (growth and selection) might balance and
influenced by the environment optimally tune the “size”
of the networks involved in sensori-motor as well as
cognitive processes.
Future work is directed toward a more formal
definition of the many facets involved in the
developmental paradigm. It is worth mentioning that
the sensori-motor layer represents only a very first stage
of development, which is at the moment artificially
carried out without cognition. This is clearly unrealistic
(in human development); a value system, drives and
other higher-level processes are certainly needed to
complement the model and give the robot the abilities
to bootstrap autonomously to new levels of
development.
References
Atkinson, J. (1998). The 'Where and What' or 'Who and
How' of Visual Development. In F. Simion &
G. Butterworth (Eds.), The Development of
Sensory, Motor and Cognitive Capacities in
Early Infancy (pp. 3-24). Hove, East Sussex:
Psychology Press Ltd.
Bertenthal, B., & Von Hofsten, C. (1998). Eye, Head
and Trunk Control: the Foundation for Manual
Development. Neuroscience and Behavioral
Reviews, 22(4), 515-520.
Brooks, R. (1996, November 4-8). Behavior-Based
Humanoid Robotics. Paper presented at the
IEEE/RSJ IROS'96, Osaka, Japan.
Di Pellegrino, G., Fadiga, L., Fogassi, L., Gallese, V.,
& Rizzolatti, G. (1992). Understanding motor
events: a neurophysiological study.
Experimental Brain Research(91), 176-180.
Edelman, G. M. (1988). Neural Darwinism: The Theory
of Neuronal Group Selection. Oxford: Oxford
University Press.
Hirai, K., Hirose, M., Haikawa, Y., & Takenaka, T.
(1998, May, 16-20). The development of
Honda humanoid robot. Paper presented at the
IEEE International Conference on Robotics
and Automation, Leuven, Belgium.
Johnson, M. H. (1997). Developmental Cognitive
Neuroscience ( 3 ed. Vol. 1). Malden, MA and
Oxford UK: Balckwell Publisher Inc.
Kovacs, I. (2000). Human development of perceptual
organization. Vision Research, 40, 1301-1310.
Metta, G. (2000). Babybot: a Study on Sensori-motor
Development. Unpublished Ph.D. Thesis,
University of Genova, Genova.
Metta, G., Carlevarino, A., Martinotti, R., & Sandini,
G. (2000). An Incremental Growing Neural
Network and its Application to Robot Control.
Paper presented at the International Joint
Conference on Neural Networks, Como, Italy.
Metta, G., Sandini, G., & Konczak, J. (1999). A
Developmental Approach to Visually-Guided
Reaching in Artificial Systems. Neural
Networks, 12(10), 1413-1427.
Panerai, F., Metta, G., & Sandini, G. (2000). Visuo-
inertial Stabilization in Space-variant
Binocular Systems. Robotics and Autonomous
Systems, 30(1-2), 195-214.
Piaget, J. (1936). The origin of intelligence in children
(M. Cook, Trans.). London: Interntational
University Press, Inc. and Routledge & Kegan
Paul Ltd.
Quartz, S. R., & Sejnowski, T. J. (1997). The neural
basis of cognitive development: A Psychology, 15, 357-368.
constructivist manifesto. Behavioral and Brain
Sciences(20), 537-596.
Rizzolatti, G., & Fadiga, L. (1998). Grasping objects
and grasping action meanings: the dual role of
monkey rostroventral premotor cortex (area
F5). In G. R. Bock & J. A. Goode (Eds.),
Sensory guidance of movement, Novartis
Foundation Symposium (pp. 81-103).
Chichester: Wiley and Sons.
Sandini, G., Metta, G., & Konczak, J. (1997,
November). Human Sensori-motor
Development and Artificial Systems. Paper
presented at the AIR&IHAS ’97, Japan.
Sandini, G., & Tagliasco, V. (1980). An
Anthropomorphic Retina-like Structure for
Scene Analysis. Computer Vision, Graphics
and Image Processing, 14(3), 365-372.
Thelen, E., & Smith, L. B. (1998). A Dynamic System
Approach to the Development of Cognition
and Action ( 3 ed.). Cambridge, MA: MIT
Press.
Turkewitz, G., & Kenny, P. A. (1982). Limitation on
Input as a Basis for Neural Organization and
Perceptual Development: a Preliminary
Theoretical Statement. Developmental
Vapnik, V. N. (1998). Statistical Learning Theory. New
York: Wiley.

Sound localization
Simple vision

Saccades
Linear controller

Saccades Visuo-acoustic

Simple VOR
Head
VOR

ATNR

Reaching

Time

Figure 2 The developmental progression of the Babybot. A representation of how the various phases of the robot’s
development are interrelated. The light gray blocks are the initial configuration. See text for description of the
learning blocks.
 Robots As Cognitive Tools
Information theoretic analysis of sensory-motor data
Max Lungarella (lunga@ifi.unizh.ch)
Rolf Pfeifer (pfeifer@ifi.unizh.ch)
Artificial Intelligence Laboratory, Dept. of Information Technology
University of Zurich, Switzerland
Abstract
In this paper we explore the possibility of employing
information theoretic measures, like the Shannon
Entropy and the Mutual Information, in order to more
quantitatively describe the sensory data and the
interactions between the different sensory channels of an
agent as it interacts with the real world. We motivate the
use of Shannon’s entropy and of mutual information,
two fundamental concepts from information theory, as
alternative statistical tools. Entropy accounts for the
diversity or variability of a sensory or a motor signal.
Mutual information, on the other side, helps us
understand the amount of interaction between two or
more sensory channels. The results of various
experiments are presented. They show that an
appropriate sensory -motor coordinated interaction can
lead to a regularization of the sensory data. We hope
that the insights emerging from this research will
eventually lead to a more formal description of some of
the design principles of autonomous agents.
Keywords: Information theory, sensory -motor
coordination, developmental robotics, principle of
ecological balance
Introduction
Over the last decade-or-so, a number of researchers
have proposed a developmental perspective on artificial
intelligence and robotics. The ultimate shared goal
among them seems to be the idea of bootstrapping high-
level cognition through a process in which an agent
interacts with a real physical environment over
extended periods of time. In this paper we will only
consider embodied approaches, i.e., approaches taking
into account the interaction of a physical system with
the real world. An early attempt was the Cog project by
Brooks et al. (1998) at the MIT Artificial Intelligence
Laboratory, a more recent one the Babybot project at
the LIRA Lab in Genova (Metta, 1999). Kuniyoshi
(1998) and Asada (2000) have been working in a
similar direction. Some of this research is focused on
robot construction, some of it is rather metaphorical,
and some is focusing on internal mechanisms.
The present paper represents an attempt at a more
quantitative approach to the development of embodied
cognition. The goal is to acquire an understanding of
the sensory data that is generated in the different
sensory channels as a result of an agent’s interaction
with the real world, as this is, so to speak, the “raw
material” the neural substrate has to process. The way
this data is generated depends, in turn, on the particular
embodiment of the agent, i.e., its morphology (type,
position, and characteristics of the individual sensors,
and actuators), as well as on the materials involved.
In previous work (Pfeifer and Scheier, 1997, 1998;
Scheier and Pfeifer, 1997; Scheier et al., 1998) we have
been focusing on categorization, i.e., the ability to make
distinctions in the real world, since categorization is
one of the most fundamental cognitive abilities. A
natural organism that cannot make distinctions will
neither have much of a chance of survival nor will it be
of any use if it is a robot. In this research,
categorization was implemented as a process of
sensory -motor coordination as suggested early on by
John Dewey in 1896(!), later by Gerald Edelman (1987)
and Esther Thelen and Linda Smith (1994). This
approach was chosen to overcome the problems of
classical – disembodied – categorization models like
ALCOVE (Kruschke, 1992), which view categorization
as a process of mapping an input vector onto category
nodes. Classical models start from the assumption that
there is an input vector consisting of “psychologically
relevant dimensions”, whereas an agent interacting with
the real world is exposed to a continuously varying
stream of sensory stimulation, which represents a
completely different problem. It has been shown for
simple cases that through sensory-motor coordination a
temporarily stable pattern of sensory stimulation can be
induced, and a dimensionality reduction of the high-
dimensional sensory space can be achieved (e.g. Pfeifer
and Scheier, 1998; Lungarella et al., 2001) (in
ALCOVE, for example, there are typically three to four
nodes in the input layer, which constitutes only a low-
dimensional space). If done properly, sensory-motor
coordination will lead to the generation of “good”
sensory data, i.e. data which can result in stable
categorization behavior.
The present paper explores the possibility to employ
information theoretic measures to more quantitatively
describe the sensory data and the interrelation between
the different sensory channels as the agent interacts
with the real world. We start with a short discussion of
some basic aspects of sensory -motor coordination.
Then we describe a number of exp eriments and analyze
the results. Finally we discuss what we can learn from
this type of analysis and point to some future work.
Sensory-motor coordination
By definition sensory-motor coordination involves both
the sensory and the motor systems, in other words, it
involves the agent’s body. As mentioned above, the
sensory stimulation that the neural system has to
process depends on the agent’s morphology, and its
behavior: through its actions, in particular through
sensory -motor coordination, an agent can induce stable
sensory patterns in different sensory channels that can
be exploited to form cross-modal associations (Pfeifer
and Scheier, 1997). The cross-modal associations seem
to a basic prerequisite for concept formation (Thelen
and Smith, 1994), which in turn are of fundamental
importance for the emergence of what might be called
high-level cognition. Cross-modal associations, which
also depend on the agent’s morphology, nicely
demonstrate how embodiment does not only have
physical implications, but also information theoretic
ones. So, the sensory stimulation is influenced by at
least two closely related factors, morphology, and
sensory -motor coordination.
Exploration strategies are particular instances of
sensory -motor coordinations, used to extract different
kind of information from the surrounding environment.
Tactile information picked up in combination with
systematic exploratory movement of the hand (or
mouth), yields richer sensory stimulation and thus
potentially more and better information than passive
contact, e.g., particular hand movements that can be
identified as being critical to the ability to recognize
objects. Results from studies on human subjects
indicate that people explore objects consistently using
different exploratory hand movements, depending on
the knowledge (information) they are instructed to
obtain (Lederman and Klatzky, 1990). Particular
exploration procedures are used to extract hardness,
pressure, texture, or compliance, because they provide
the sensory input which is “optimal and sometimes
necessary”, for extracting the desired information.
The same holds for vision. Eye movements, for
instance, have a task-dependent character, and very
much depend on what perceptual judgment we are
asked to make (Yarbus, 1967). Differences between
tasks influence the way we pick up information, which
may or may not maximize the information intake. In
other words, eye movements influence the statistics of
the effective visual input. (Lee and Yu, 1999) sketch an
active perception framework based on information
maximization to reason about the organization of
saccadic eye movements.
Experimental setup and experiments
Our previous work on categorization which lead to the
hypothesis of dimensionality reduction through
sensory -motor coordination is suggestive, but is only
based on very simple cases. To further corroborate this
hypothesis we increased the complexity of the agent.
According to the “principle of ecological balance”
(Pfeifer, 1996; Pfeifer and Scheier, 1999), in order to
achieve mo re interesting kinds of sensory-motor
coordination, there has to be a balance of the
complexity among the agent’s task environment and the
sensory, motor, and neural system, which makes also
sense from a developmental perspective. Bushnell and
Boudreau (1993) talk about “motor development in the
mind”, i.e., in human infants there is a co-development
of the sensory and motor system.
To be able to simulate this co-development, our
experiments are performed using a five degrees of
freedom industrial robot manipulator, equipped with a
color CCD-camera, that is mounted on the robot’s end-
effector. This setup is often called an eye-in-hand
configuration (see fig. 1). The camera is the only
exteroceptive sensor used in this set of experiments.
Video frames are recorded at a rate of 10Hz, and the
resolution is reduced to 192x192 pixels. The control of
the robot is image-based, which implies that there is
visuo-motor couipling, in other words the desired end-
effector position is achieved by processing in an
adequate way the camera image. The experiments are
performed in an unstructured, but static real world
environment, cluttered with a variable number of
colored objects, of different form, color and size.
The robot’s task is to foveate on red-colored objects,
independently of their shape, and size. The control
architecture is hard-wired. The sensory part consists of
two one-dimensional arrays of color-sensitive cells
arrange as a cross, in the following referred to as “1D
retinas”. The 1D retinas consist of a certain number of
color sensitive rectangles (which might be interpreted
as receptors or sensory channels) – there are M
receptors for the horizontal retina, and N for the vertical
one. The output of an individual receptor is the average
over a rectangular patch of pixels. The receptor density
is variable and depends on the horizontal or vertical
position, x or y, respectively, i.e., δR ( x) , and δR(y) for
the red-receptor, for instance - see figure 2. In this
paper three receptor types are considered: red (r), green
(g), and blue (b). An additional receptor, sensitive to
r+g+b
intensity, is obtained as I = . An attenuation
3
of the influence of changing lighting conditions is
achieved through a color-space transformation - see
also (Itti et al., 1998). Three “broadly” color-tuned
g +b
channels are created: R = r − for red,
2
r +b r+g
G= g− for green, and B = b − for blue.
2 2
The negative values are set to zero. Each channel yields
maximal response for the pure, fully-saturated color to
which it is tuned, and yields zero response for black
(r=g=b=0) and white (r=g=b=255) inputs.
Figure 1: Basic manipulator geometry
On the motor side (refer to figure 1), the motoneuron
controlling joint J0 (shoulder), which is responsible for An additional feature of the control architecture is a
the rotation around the vertical axis, is fully connected habituation (or boredom) coefficient h. Its purpose is to
to the color-space transformed outputs of the horizontal avoid situations in which the robot keeps focusing on
1D-retina. The motoneurons of joint J1 (shoulder), and one and the same object. Its effect is to move the robot
joint J4 (wrist), which are responsible for the up and into random joint angle configurations.
down movement, and which are “kinematically”
The sensory -motor coordinated behavior is compared
coupled (see eqn. 1), in order to keep the camera always
with one that is not sensory-motor coordinated, where
horizontal, are fully connected to the receptors of the
vertical retina. Furthermore, once the robot has the joints are actuated randomly.
successfully foveated on a red object, joint J2 (elbow) is
actuated in an oscillatory manner - inducing a back and
forth movement of the end-effector. Its purpose is a
slightly more complex sensory-motor coordinated
interaction with the object itself. The weights of the
various neural connections are hard-wired, and can be
thought of as some sort of basic motor reflexes. The
control architecture is a “Braitenberg-style” reactive
architecture, with direct sensory-motor couplings. The
equations employed for updating the joint-angles Ji [n]
are (eqn. 1):

 J 0 [n + 1] = J 0 [n ] + f 0[n ] ⋅ M ∑ w0 i ⋅ RH i[n ]
 i= 0
 N
 J 1[n + 1] = J 1[n ] + f 1[n ] ⋅ ∑ w1 j ⋅ RV j[n ]
 j =0

 J 2 [n + 1] = J 2 [n ] + f 2 [n ] ⋅ ( w0 , 2 + w0 , 2 )
N ( N −1)
Figure 2: The camera-image is rectangular. The two
 [
 J 4 n + 1] = J 4 [n ] + 90 − ( J 1[n ] + J 2[n ]) one-dimensional retinas (horizontal and vertical) are
shown.
where f0 [n]=c0 (which is a constant) if J0 [n]<J0min ,
Analysis methods
f0 [n]=-c0 , if J0 [n]>J0max, and f0 [n]=f0 [n-1] else -
identical equations hold for f1 [n], f2 [n], J1 [n], and As mentioned above we are interested in the
J2 [n], respectively. w0i and w1j represent the weights information theoretic implications of embodiment.
that connect the output of the horizontal and vertical More specifically we want to investigate the use of
retinas to the motoneurons. information theoretic measures (Shannon, 1948; Cover
and Thomas, 1991; Papoulis, 1991) like the Shannon
entropy H, the Shannon or mutual information MI, and
its normalized companion MI , if applied to the sensory
channels of a situated autonomous agent. Many of the
results of information theory are in terms of these
fundamental quantities. The Shannon entropy H(X)
accounts for the potential diversity or variability that is
displayed by the random variable X (see appendix for a
definition), where X represents the signal from the
sensory channel. This is in principle equivalent to our
intuitive notion of information – the more unpredictable
a signal or an event, the more information its
occurrence contributes. MI ( X ; Y ) is a measure that
takes into account linear as well as nonlinear
dependencies between two time series (observations of
a stochastic process). This is in contrast to the better
known correlation function, which measures just linear
dependencies.
The application of information theory is not devoid of
problems. One of the greatest problems is the huge data
requirement. To avoid it, (strong) assumptions about
the signals involved, and/or the noise have to be made.
These assumptions are often unfounded and difficult to
test. Nevertheless, assuming true independence of two
random processes, or normality of the signals, can
significantly cut down the number of measurements
required for the analysis. In our case no such
assumptions about the sensory and motor data are
necessary, because a sufficiently large number of data
has been collected The number of samples per
experiment is around 3000, sufficient for the
computation of measures like entropy and mutual
information.
A straightforward method of computing entropy and
mutual information is to estimate the first and second
order probability density functions - p(x) and p(x,y),
which is done by normalization of the 1st and 2nd order
histograms of the time-series. The entropy (or self-
N respectively, and compute MI ( X ; Y ) . For perfectly
information) is given by H ( X ) = − ∑ p ( xi ) log p ( xi ) ,
i =1
for a discrete variable xi , which can be in N possible
states {x1 ,x2 ,…xN}. The mutual information is defined as
MI(X;Y)=H(X)+H(Y)-H(X;Y), where
NM
H ( X ; Y ) = − ∑ ∑ p ( x , y ) log p ( x, y) is the joint
entropy – the discrete random variable X and Y have N
or M different states, respectively. Furthermore we
define the redundancy as the “capacity-normalized”
MI ( X ; Y )
mutual information MI ( X ; Y ) = , with
C bumps in the center of the 3D plot. In other words, the
C = max MI ( X ; Y ) , where the maximum is taken over
p ( x)
all possible densities p(x). In our case
C = max H ( X ) =8bit for all sensory channels, i.e., the
p( x )
sensory signals assume discrete values between 0 and
255=28 -1. Usually the channel capacity C is measured
in bits/sec, and the rate of transmitted information
equals the entropy rate H ( X ) bits/sec. We normalize
everything by fs , which is the rate at which the sensors
are read out. These measures are applied to the four
previously introduced sensory sub-modalities: red (R),
R+G+B
green (G), blue (B), and intensity ( I = ).
3
Results
We performed four different kinds of analyses.
The Shannon entropy (displayed on the vertical axis) of
the red, green, and blue sensory channels of the
horizontal linear resolution retina in the case of a
random movement can be seen on the left side of figure
4. The retina is composed of 18 color-tuned receptors
(R, G, and B-channels). The graph on the right shows
the case of sensory -motor coordinated behavior
(foveation on red-colored objects). In the central part of
the retina (from receptor 8 to receptor 12) the
information inflow through the R-channels is increased,
whereas the one in the peripheral part is decreased, if
compared to the not sensory-motor coordinated case.
The behavior of the G-, and B-channels is reversed, i.e.,
the flow through the central channels is decreased. The
same holds for the vertical linear resolution retina,
where the effect is less pronounced, but still visible (not
shown here). The graphs are averages over six
experimental runs.
The intrasensory information overlap between two R-
channels is shown in figure 5. Basically we set X=Ri
and Y=Rj , where Ri and Rj are the ith , and jth R-channel
dependent channels MI ( X ; Y ) =1, while for completely
independent ones MI ( X ; Y ) =0. For not sensory-motor
coordinated interaction, in our case the result of a
random actuation of the joints, the information overlap
is minimal, there is no redundancy. Remember that
H (X)
MI ( X ; X ) = , i.e., the redundancy is equivalent
C
to the normalized entropy or self-information, which
explains the diagonal ridge. The case of sensory-motor
coordinated interaction is shown on the right side of
figure 5. The information overlap is evident from the
amount of information shared by the R-channels in the
foveal part of the 1D-retina, is larger than the one of the
R-channels in the peripheral part. This redundancy is a
consequence of the interaction itself.
Figure 6 shows the information overlap between two
sub-modalities of the same sensor modality – color (R)
and intensity ( I ). The color channel and the intensity
channel share the same spatial location, but are tuned
to different stimulus dimensions, color and intensity,
respectively. Again, as for the Shannon entropy, if
compared to the not sensory-motor coordinated case,
the result of the sensory-motor coordination is an
increase of the mutual information in the central part of
the retina, and a decrease in the peripheral part.
The amount of variability (information) must not be
confused with the cumulated amount of activation (total
stimulation) of a particular sensor. Figure 7 illustrates
this point. The not sensory-motor coordinated case can
be seen on the left side, whereas the sensory-motor
coordinated case on the right side. Since the robot’s
task is to foveate on red objects, the peak in the
cumulated stimulation of the R-receptor is obvious –
the agent spends a lot of time foveating on red objects.
Less obvious is the fact that the maximum of the
entropy of the B-receptor for the same dataset is
slightly larger than the maximum for the R-receptor –
see figure 4. More striking is the difference for the not
sensory -motor coordinated case, where the relationship
is even reversed! The entropy of the G-, and B-channel
has an average of 5.7bit, the one of the R-channel
4.7bit. The cumulated stimulation is higher for the R-
channel, though.
Figure 4: The Shannon Entropy for different sensory
channels measured in bits. No sensory-motor
coordination (random) on the left, sensory-motor
coordination (foveation on red objects) on the right.
Figure 5: Mutual information between receptors of the
same sensor modality. Random actuation on the left.
Sensory-motor coordination on the right.
Figure 6: Mutual information between R-,G-,B-
channels and I-channel (intensity) for different sensory
channels. Random movement on the left. Foveation on
red objects on the right.
Figure 7: Cumulated stimulation of the R-,G-,B-
receptors. The sensory-motor coordinated case is on the
right.
Discussion
The following few points are of interest. First of all, the
amount of information flowing through a particular
sensory channel is increased/decreased through an
appropriate sensory-motor coordinated interaction. An
upper bound is given by the channel capacity, i.e.,
H=C. Statistical information is maximized when all
possible sensory states of the discretized signal have
equal probability of occurrence, which means that the
probability density function (pdf) is uniform. A lower
bound is given by H=0, no variability at all (constant
sensor stimulation), i.e., a pdf with a single peak.
Exploration strategies, that lead to a balance between
predictability (low entropy) and variability (high
entropy) of the sensory signal need further
investigation. In communication theory the goal is not
quite the same. Given a certain amount of noise, the
information transfer through the communication
channel has to be maximized, the more information can
be pushed through the channel, the better.
The second point of interest is that sensory-motor
coordinated interaction leads to redundancy in the
sensory channels of the same and of different
modalities, i.e., to a higher mutual predictability
between them. If the two signals are totally
 H ( X ) + H (Y ) − H ( X ; Y )
uncorrelated, MI ( X ; Y ) =
C We may also make a step forward in describing the
the joint entropy equals the sum of the individual
entropies. The same measure reaches its maximum if
the entropies of the individual sensory channels are
high, and there is a high correlation among them (low
joint entropy) – one sensory channel can be used as a
predictor for the other one. This redundancy is clearly
not present in the case of not sensory -motor coordinated
interaction. The mutual predictability in this case is
much lower – see figure 5.
Human infants exhibit a wide range of exploration
strategies: mouthing, banging, fingering, scratching,
squeezing, waving, and listening (Kellman and
Arterberry, 1998). When objects are placed in the
mouth, infants are able to detect surface properties
(Meltzoff and Borton, 1979), and object characteristics
such as rigidity (Rochat, 1987). In other words, there
are different actions related to the exploration of
different object properties, in the sense, that they
provide the sensory input, which is “optimal and
sometimes necessary”, for extracting the desired
information. As shown in this simple case study, the
information inflow through the various sensory
channels very much depends on the action itself, i.e., an
appropriate choice of it seems to be of importance for
the simplification of the subsequent neural processing.
Since entropy is an information theoretic measure that
captures the variability of the (sensory and motor)
signals, it tells us something about the complexity of
the interaction itself. In an agent context, the more
diverse the agent’s behavior, the more variation in the
sensory channels, and the higher the entropy in the
sensory system. Since we are interested not only in
complexity due to sensory stimulation, but also in
complexity due to self-generated sensory stimulation,
we need to take into account aspects of the motor
system’s variability. A good measure to start with, is
the ratio between the total entropy of the motor signals
x=(x1 ,…, xS ) and the total entropy of the sensory signals
Hmotor ( x )
y=(y1 ,…,yT), we define it as B =
Hsensor( y )
should be a match in the variability in the sensory
channels and of the motor outputs. In other words, B
measures how well-balanced the motor and sensor
signals are.
Conclusion and future work
With the ideas exposed in this paper, it may eventually
be possible to more formally describe some of the
design principles of autonomous agents (Pfeifer, 1996;
Pfeifer and Scheier, 1999), e.g., the principle of
ecological balance mentioned earlier.
=0 , and
requirements for adaptive agents: The individual
components (dimensions in sensory and motor space)
must have a lot of variability, but they must also be able
to couple to others for specific tasks. This is precisely
what complex systems are about.
The importance of an appropriate sensory-motor
coordinated interaction cannot be overestimated, since,
as the results show, it can lead to a regularization of the
raw sensory stimulation – see also (Pfeifer and Scheier,
1999) and (Lungarella et al., 2001) - which in turns
should speed up and simplify learning. We hypothesize
that the self-information (entropy) of the central and
peripheral receptors largely depends on the type of
interaction, but is less dependent on the environment. In
a similar way eye-movements, and certain particular
hand movements have task-dependent characters.
Many additional experiments need to be performed to
confirm our hypotheses, though. Other sensory
modalities (touch, audition) have to be taken into
account. They would shed some light on how infant
discover intermodal relationships, and how the
existence of multiple sensory channels allows us to
learn more about, and function within, the world.
Different sensor morphologies and other task
environments have to be tested and they’re effect on the
information inflow need to be explored.
The next step will then be to exploit the data for
learning. Using similar kinds of analyses, we hope that
we will be able to derive the conditions under which
agents can learn rapidly new categorization behavior
while maintaining the stability of existing ones.
Acknowledgments
We would like to thank Lukas Lichtensteiger, who
provided us with valuable suggestions and helpful last-
minute comments. Furthermore we are indebted to
Fumiya Iida for the many discussions, which helped to
. There shape the ideas exposed in this paper. This research has
been supported in part by grant #11-57267.99 of the
Swiss National Science Foundation.
Information Theoretic Appendix
Some useful definitions:
- Random variable, RV: variable that assumes a
numerical value for each random outcome of an event
or experiment
- Probability density function of a RV X, p(x):
normalized 1st -order histogram of a RV.
 - Joint probability density function of RV X and Y,
p(x,y): normalized 2nd -order histogram of 2 RVs.
- Shannon entropy H(X): measures the randomness of
a RV. The more random a RV, the more entropy does
it have. Intuitively it is a measure of (the logarithm
of) the number of states the RV could be in.
- Joint entropy H(X;Y): Measures the uncertainty
about both X and Y.
- Mutual Information MI(X;Y)=H(X)+H(Y)-H(X;Y):
Measures the portion of entropy shared by X and Y.
It is high if both X and Y have high entropy (high
variance), and share a large fraction of it (high
covariance). It is zero, if X and Y are statistically
independent. In other words, MI is a measure of the
deviation from statistical independence.
- Channel capacity C= max H ( X ) : In our case the
p ( x)
maximal amount of statistical information that can be
transferred through a sensory channel in a certain
instant. It is computed over all possible sensory signal
distributions p(x).
- Redundancy MI ( X ; Y ) = MI ( X ; Y ) / C : The
capacity- normalized mutual information.
Since the logarithm in base 2 is used, H and MI are
measure in bits. Both, entropy and mutual information
are used in a statistical connotation; they can be thought
of as multivariate generalizations of variance and
covariance (univariate statistics) that are sensitive to
both linear and nonlinear interactions.
References
Asada, M., MacDorman, K.F., Ishiguro,H., and
Kuniyoshi, Y. (2000). Cognitive developmenal
robotics as a new paradigm for the design of
humanoid robots. Proc. Of the 1st IEEE-RAS Int.
Conf. on Humanoid Robotics, MIT, Cambridge, MA.
Braitenberg, V. (1984). Vehicles: An Experiment in
Synthetic Methodology. Cambridge, MA: MIT Press.
Brooks, A.R., Breazeal, C., Marjanovic,M., Scassellatti,
B., and Williamson, M. (1999). Alternate essences of
intelligence. AAAI-98.
Bushnell, E.M., and Boudreau, J.P. (1993). Motor
development in the mind: The potential role of motor
abilities as a determinant of aspects of perceptual
development. Child Development, 64, 1005-1021.
Cover, T.M., and Thomas, J.A. (1991). Elements of
Information Theory. New York: Wiley.
Edelman, G.E. (1987). Neural Darwinism: The Theory
of Neural Group Selection. New York: Basic Books.
Itti, L., Koch, C., and Niebur, E. (1998). A model of
saliency-based visual-attention for rapid scene
analysis. IEEE Transactions on Pattern Analysis and
Machine Intelligence, 20(11), 1254-1259.
Kellman, P.J., and Arterberry, M.E. (1998). The Cradle
of Knowledge. Cambridge, MA: MIT Press. A
Bradford Book.
Kruschke, J.K. (1992). ALCOVE: An exemplar-based
connectionist model of category learning.
Psychological Review, 99, 22-44.
Kuniyoshi, Y., and Berthouze, L. (1998). Neural
learning of embodied interaction dynamics. Neural
Networks, 11, 1259-1276.
Metta, G. (1999). Babyrobot: A Study on Sensori-motor
Development. In Dipartimento di Informatica,
Sistemica e Telematica. University of Genoa (PhD
Thesis). Genova. Italy.
Lederman, S.J., and Klatzky, R.L. (1990). Haptic
exploration and object representation. In M. Goodale
(ed.), Vision and Action: The Control of Grasping.
pp. 98-109. New Jersey: Ablex.
Lee, T.S., and Yu S.X. (1999). An information-
theoretic framework for understanding saccadic
behaviors. In Solla, S.A., and Leen, T.K. (Eds.).
Proc. of the 1 st Conf. on Neural Information
Processing. K-R. Muller. MIT Press.
Lungarella M., te Boekhorst R., and Pfeifer, R. (2001).
Dimensionality reduction through sensory-motor
coordinated interaction. In preparation.
Papoulis, A. (1991). Probability, Random Variables,
and Stochastic Processes. McGraw-Hill Book
Company, NY, USA.
Pfeifer, R. (1996). Building “Fungus Eaters”: Design
principles of autonomous agents. In P.Maes,
M.Mataric, J.-A.Meyer, J.Pollack, and S.W.Wilson
(Eds.). From animals to animats: Proc. of the 4th Int.
Conf. on Simulation and Adaptive Behavior.
Cambridge, MA: MIT Press. A Bradford Book.
Pfeifer, R., and Scheier, C. (1997). Sensory-motor
coordination: the metaphor and beyond. Robotics and
Autonomous Systems, 20, 157-178.
Pfeifer, R., and Scheier, C. (1998). Representation in
natural and artificial agents: a perspective from
embodied cognitive science. Zeitschr. f.
Naturforschung, 53c, 480-503 (invited paper).
Pfeifer, R., and Scheier, C. (1999). Understanding
Intelligence. Cambridge, MA: MIT Press.
Rochat, P. (1987). Mouthing and grasping in neonates.
Evidence for the early detection of what hard and soft
substances afford for action. Infant Behavior and
 Development, 25, 871-884.
Scheier, C., and Pfeifer, R. (1997). Information
theoretic implications of embodiment for neural
network learning. Proc. ICANN 97, 691-696.
Scheier, C., Pfeifer, R., and Kuniyoshi, Y. (1998).
Embedded neural networks: exploiting constraints.
Neural Networks, 11, 1551-1569 (invited paper).
Shannon, C.E., (1948). A mathematical theory of
communicaton. Bell Syst. Tech. J., 27, 379-423, 623-
653.
Thelen, E., and Smith, L. (1994). A Dynamic Systems
Approach to the Development of Cognition and
Action. Cambridge, MA: MIT Press. A Bradford
Book.
Yarbus, A.L. (1967). Eye Movements and Vision. New
York: Plenum Press.
 The Constitution of Spatiality in Relation to the Lived Body :
a Study based on Prosthetic Perception
Charles Lenay (charles.lenay@utc.fr),
COSTECH, Université de Technologie de Compiègne,
Centre P.Guillaumat, BP 60319 – 60206 Compiègne Cédex France

Olivier Gapenne (olivier.gapenne@utc.fr)

COSTECH, Université de Technologie de Compiègne,
Centre P.Guillaumat, BP 60319 – 60206 Compiègne Cédex France

John Stewart (john.stewart@utc.fr)

COSTECH, Université de Technologie de Compiègne,
Centre P.Guillaumat, BP 60319 – 60206 Compiègne Cédex France

Abstract simple substitution of an entry via the optic nerve by a

Studies using the sensory substitution devices designed
for visually impaired persons reveal that perceptive
activity itself is embodied in a lived body capable of
movement and possessing its own spatial dimensions.
We have used these prosthetic devices, based on the
substitution of the visual sensory input by a tactile
sensory input, in order to carry out a systematic study of
perception, and in particular of spatial depth. We show
that the spatial localisation of a target requires dynamic
sensori-motor coupling, and that this activity involves
the spatial dimensions of the lived body of the
perceiving subject.
Introduction
We have been engaged for several years in a
fundamental research programme on the functioning
and appropriation of technical devices for
supplementing perceptive capacities (Lenay, 1997 ;
2000 ; Hanneton, 1999). This programme is based on
an extension of pioneering work of Bach-y-Rita on
"sensory substitution" systems. These devices have
been developed, since the end of the 60's, in order to
help persons with congenital or acquired visual
impairment (Bach y Rita, 1994 ; Kaczmarek, 1995).
The first "Tactile Vision Substitution System" (TVSS)
converted a visual image captured by a video camera
into a tactile "image" composed of a 20 x 20 array of
stimulators placed either on the back or on the chest
(Collins, 1973). The initial studies with this device
produced three results which are quite fundamental,
and whose validity and relevance has only been
confirmed and amplified by subsequent research :
i) Firstly, the presentation of shapes to an immobile
camera only permits a very limited discrimination of
the received stimuli, and the latter are perceived as
being situated on the surface of the skin. Thus, the
tactile entry does not, in itself, give access to spatial
perception.
ii) However, if the subject disposes of the capacity to
actively manipulate the camera (movements from left to
right, up and down, zoom), he or she develops a
spectacular capacity to recognize shapes. The first step
is learning how variations in sensation follow from
actions : when the camera is moved from left to right,
the stimuli on the skin shift from right to left; when
zooming in, the stimuli expand, etc. When the subject
has learned to aim the camera at a target, he or she
begins to discriminate lines and surfaces, and then to
recognize familiar objects of increasing complexity to
the point of being able to discriminate faces.
iii) Thirdly, this recognition is accompanied by a
"projection" of the percepts into objects with external
spatial localisation. Initially, the subject feels the
successive stimulations on the surface of the skin.
However, as perceptual learning proceeds, these
sensations of touch fade from consciousness, and are
replaced by the perception of stable objects, situated
"out there" in front of the subject. Thus, according to
the testimony of the subjects, the proximal irritations
produced by the tactile display unit are quite distinct
from the perception itself. This subjective localisation
of objects in space comes about quite rapidly, after 5 -
15 hours of familiarisation and learning. Congenitally
blind subjects discover perceptual concepts which are
radically new for them, such as parallax, shadows, the
interposition of objects, etc. Certain classical optical
illusions are spontaneously reproduced (Bach y Rita,
1982 ; Guarniero 1977). These experiments can be
performed not only with visually impaired subjects, but
also by blindfolded normal subjects.
The perceptual learning involved in these studies
reveals the astonishing plasticity of the central nervous
system, which undergoes vast functional
reorganisation. The tactile sensory input has no
intrinsic relation to the retinal input of the visual
system, and the motor control of the camera by the
hands has no intrinsic relation to the motor command of
the ocular muscles. Nevertheless, the brain exhibits the
capacity to constitute a perceptual world composed of
forms and events which correspond to those given to us
by visual perception. In addition, if the tactile display
unit is displaced form the chest to the back, and the
hand-held camera is replaced by a miniature camera
attached to the frame of a pair of "spectacles", a
practised subject is able to adapt almost immediately. In
a few seconds, the subject recovers a distal perception
"out there" in front of him.
The essential role of action in the progressive
emergence of structured percepts lends support to the
hypotheses of active perception. In this perspective, we
abandon the conception of perception as a process in
which the system passively receives an input in the
form of "information", and then performs
computational operations in order identify objects and
events in the form of internal representations. On the
contrary, it is by means of his own actions that the
subject seeks and constructs the regularities in the
relations between action and sensation. What is
perceived and recognized is not so much invariants of
sensation, but rather invariants in the circular sensori-
motor loops that are inseparable from the activity of the
subject (Gibson, 1966 ; 1986 ; Paillard, 1991 ; Varela,
1979 ; Turvey, 1995 ; Edelman, 1987 ; O’Regan,
2001). It follows that richness of perception depends as
much (if not more) on the qualities of action (mobility,
rapidity, and indeed the whole range of actions that are
qualitatively possible) than on the qualities of sensation
(spectral range, the number of sensors, etc).
The requirement that the subject should be able to
actively manipulate the sensory captors means that the
phrase "sensory substitution" is, strictly speaking,
inappropriate. It is not sufficient for a technical device
to give access to novel sensory inputs; it must also
provide opportunities for novel actions, with
appropriate sensory consequences.
This type of prosthetic device, in addition to its purely
practical import, offers an original experimental
situation for the study of perceptive activity. It opens
up the possibility of an empirical access to the
consciousness of objects as situated externally in space.
Specifically, it is possible to experimentally reproduce
the genesis of depth perception in adult subjects, and
hence to observe the processes involved.
Materials and methods
Our method of investigation consists of a deliberate
simplification of the sensori-motor coupling, reducing it
to the bare minimum that is still sufficient for a subject
to perceive an object as localized in external space
(Marque, 2000). The sensory input, in particular, is
reduced to the simplest form possible, a single
vibratory tactile stimulator that is either "on" or "off".
The stimulator is held in one hand, and activated by a
simple photoelectric cell fixed on the index finger of
the other hand. The photoelectric cell captures incident
light in a fairly wide angle (about 20°); it is activated
(and hence activates the tactile stimulator) when the
light intensity is greater than a certain threshold.
target
Photo-electric cell
Index
tactile stimulator
A luminous target is placed at a certain distance from
the subject in a dark room. Subject are blindfolded, and
can freely move their arm, hand and finger to which the
photoelectric cell is attached. After several minutes
exploration, the subjects (20 sighted and 10 unsighted)
are able to localize the target. There is two
experimental conditions, distance and direction, of 6
target’s position for a pointing task. We take the
average distance between finger’s pointing and target’s
position. Subjects were able to distinguish between the
six targets as it is shown by the confidence ellipses or
by verbalisation (on the basis of a series of reference
positions previously learned) (Hardy 2000).
To sum up: when the subject has succeeded in "latching
on" to the target, he is rapidly able to indicate its
position, both in respect to direction and in depth.
It will be noted that under these conditions, with just a
single point of tactile stimulation, corresponding to a
single receptive field, the sensory input is reduced to a
binary temporal sequence: "off....... on.. off... on... off"
etc; in other words, the sensory input as such contains
absolutely no possible reference to any spatiality. We
may recall that in the TVSS, there were 400 points of
stimulation, arranged in a 20 x 20 array that
corresponded to the receptive fields of the video
camera. Under those conditions, it was not possible to
exclude that the input information already contained a
certain amount of characteristically spatial information.
The present experimental set-up is designed specifically
to totally exclude such an eventuality; if the target is
localized, it can only be due to active exploration. We
may say that the set-up forces a spatial and temporal
deployment of the perceptive activity; and
consequently, this activity takes the form of observable
movements that can be studied in a purely behaviourist
manner.

Analysis of the results
It is not difficult to understand how localization of the
target is possible. In fact, localization is still possible
even if the movements are themselves simplified, and
we reduce them just to movements of the arm around
the shoulder articulation, and movements of the wrist
(ie the elbow is blocked in a straight-arm position). For
convenience, in the following diagrams we consider
movements only in a 2-D horizontal plane (we recover
3-dimensional space, as in the actual experiments, if we
consider also up-and-down movements). In Figure 1,
we represent the situation in (x, y) coordinates. The
subject is situated at the origin, (0,0), which we
designate as point O. The target is a point source, S,
situated at a distance L from the subject with co-
ordinates (0,L). Point P represents the wrist of the
subject; the co-ordinates are (b.cos α, b.sin α), where b
is length of the arm, and α = (Ox, OP) indicates the
direction of the arm. The angle at the wrist, between the
arm and the hand, is designated by β = (PO, PS).
y
target
S wrist angle in order to point the finger at the target.
L
photo-electric cell
w rist P β
b
α
0 x
1 faculty of "pointing the finger" at any given point in
Figure 1: Specification of the variables L (distance of
the target), b (length of the arm), α (direction of the
arm) and β (angle at the wrist).
By trigonometry, from Figure 1 we obtain the
following formula for L, the distance of the target :
L = b(sin α - cos α tan(α+β)) Equation (1)
We may suppose that the length of the arm, b, is known
(we shall come back to this point in the discussion). It
will be noted that for α = π/2, β is necessarily 2π and
the distance is indeterminate. We may suppose that this
singular position (the arm straight out in front, the wrist
and finger straight in line with the arm) is used to set
the general orientation of the bust. Given this, from a
purely mathematical point of view a single pair of
values (α, β) could be sufficient to determine the
distance.
The experimental observation, however, is that one or
two "contacts" with the target are not sufficient for the
subjects to succeed in localising the target. On the
contrary, they perform regular oscillations around the
target: small wrist movements (i.e. exploratory
variation in β) in order to determine, for a given value
of α, the precise value of β that is necessary to obtain
the sensory feedback. Overall, it seems as though the
subjects seek to identify the functional relationship
between α and β which holds when sensory stimulation
is present. It is indeed possible to rewrite Equation (1)
to express α as a function of β :
β = 2π − α + Atan [ (b sinα - L)/(b cosα)] Equation (2)
This function is shown in Figure 2, for different values
of L, and it can be seen that the relationship between α
and β is indeed characteristic of L. In particular, for
small values of L, β increases sharply as a decreases
from π/2 ; as L increases, the curve flattens out1. Thus,
the proximity of the target is related to the rapidity with
which β must be increased to compensate for a given
decrease in α. We may also note at this point that if the
subject makes a sudden, large movement of the arm, he
becomes "lost", unable to rapidly discover an adequate
This confirms that the basis of depth perception is not
an isolated pair of values (α, β) but rather the
continuous function in Figure 2.
It seems likely that the relative facility with which the
experimental subjects succeed in this perceptive task
stems from the fact that the relevant rule governing
their movements, expressed mathematically by
Equation (2), is already well know to them as the
space. The same sensor-motor scheme is used (in
reverse) in order to "perceive with the finger". In other
words, the perception of the spatial position of the
target is neither more nor less than the "extraction" of a
stable rule of pointing. It follows that depth perception
does not occur in an abstract representational space, but
rather in the concrete dynamic coupling between the
subject and the environment. The position of the target
in depth is constructed as a sensori-motor invariant2.
1 It might be remarked that, from a purely mathematical point
of view, one could equally well express α as a function of β.
However, quite apart from the fact that the resulting
expression is algebraically awkward, this would correspond
to the subject cocking the wrist at a set angle β, and exploring
the arm angles α to obtain a sensory stimulation; such
behaviour is not observed in the experimental subjects.
2 Space itself can thus be thought of as the group of
transformations which make it possible to repeat the same
succession of sensations via the same succession of actions.
This is closely akin to the conception of the space of
perceptions developed by Henri Poincaré (1905 ; 1907).

350
300
Angle β (degree)
250
200
150
100
50
0 that any tool, from the moment that it is correctly
1,38
0
1,26
20
40
1,14
60
80
1,02
Angle α (degree)
Distance L
Figure 2. The curve expressing the angle β as a
function of the angle α for values of L ranging from
1.02 to 1.42 (the length of the arm, b, is taken as unity).
In another series of experiments, we constrained the
range of possible actions even further by requesting the
subject to block the wrist movements, thus keeping the
finger in a straight line with the arm. When the only
movements allowed were rotation of the arm from the
shoulder joint (i.e. variation in α but not β), the subject
was still able to indicate the direction of the target, but
depth perception was lost. Similarly, if movements at
the wrist were allowed, but the arm was fixed (ie
variation in β but not α), the subject was again able to
indicate direction, but depth perception was again
absent. In order for depth perception to arise, it was
necessary that the wrist, with a possibility of action
(variation in β), should itself be able to move in relation
to the target. This requires, specifically, a concrete
dimension of the arm which sweeps over the space.
This is well expressed by equations (1) and (2), which
include the parameter b, the length of the arm. This is
not to say that the subject performs a trigonometrical
computation based on an explicit value for b. In fact,
one might just as well say that the length of the arm is
constituted as a function of the distance of the target.
This could be expressed by rewriting equation (1) in a
mathematically equivalent form :
b = (sin α - cos α tan(α+β)) / L Equation (3)
The subject, as an organism in movement, belongs to
the space in which he is situated with respect to the
target. Perception is a form of embodied action,
inseparable from a lived body which confers the
possibility of acting in the world.
Conclusion
These experiments show clearly that there is no spatial
300-350
perception without embodied action, no depth without a
250-300 spatial dimension of the lived body which actually
200-250 constitutes the "units" in terms of which "external"
150-200 space can be perceived. In addition, we have seen how
100-150 a technical device can function as a modification of the
50-100 "lived body" which is constitutive of perceptual
0-50 activity. We may generalize this approach by remarking
grasped, can be understood as an instrument of
coupling which is integrated into the "lived body" of
the perceiving subject. For example, the white cane of
the blind person is an instrument of coupling which
renders accessible a tactile perception at the end of the
cane (i.e. where there are no nerve fibres at all). From
the moment that the tool is integrated in the perceptive
activity, it becomes "transparent" (i.e. it disappears
from consciousness). To take another everyday
example, an experienced driver "becomes one" with his
car; he perceives the road-surface under "his" tires; it is
"he" (and not the car) which grazes the kerb or runs
over some rough gravel. The driver is not conscious of
the vibrations in the car seat (as such), nor even of the
relations between his proximal actions (turning the
steering-wheel, stepping on the brake) and his
sensations (visual or tactile); his attention is focussed
on the outside, and on the events which occur in the
"world" in which he is acting. We may conclude by
saying that technical artefacts, by transforming our
modes of acting and sensing, transform our "lived
bodies" and are thus constitutive of our way of being in
the world.
Acknowledgments
We would like to thank Clotide Vanhoutte and
Catherine Marque for their valuable help in desinging
the equipment and carrying out the experiments
reported in this paper.
References
Bach-y-Rita, P. (1982). Sensory substitution in
rehabilitation. In Rehabilitation of the Neurological
Patient, L. Illis, M. Sedgwick & H. Granville (eds.);
Oxford: Blackwell Scientific Publications, pp. 361-
383.
Bach y Rita, P. (1994). Sensory substitution, volume
transmission and rehabilitation: emerging concepts.
In L. S. Illis (Ed.), Neurological Rehabilitation, (2nd
ed., pp. 457-468). Oxford: Blackwell.
Collins, C.C. and Bach y Rita, P. (1973). Transmission Van Gelder (Eds.) Mind as Motion. Cambridge, MA:
of Pictorial Information Through the Skin, Advan. MIT Press.
Biologic. Med. Phys., 14, 285-315. Varela, F. (1979). Principles of Biological Autonomy,
Edelman, G. (1987). Neural Darwinism: The Theory of New York: Elsevier.
Neuronal Group Selection. Basic Books, New York,
1987.
Gibson, J.J. (1966). The senses considered as
perceptual systems. Boston : Houghton Mifflin.
Gibson, J.J. (1986). The ecological approach to visual
perception. Hillsdale, Nj: Erlbaum.
Guarniero, G. (1977). Tactile vision : a personal view.
Viz. Impairment and Blindness, march, 125-130.
Hanneton S., Gapenne O., Genouel C., Lenay C.,
Marque C. (1999). Dynamics of Shape Recognition
Through a Minimal Visuo-Tactile Sensory
Substitution Interface, Third Int. Conf. On Cognitive
and Neural Systems, Mai 1999, Boston, USA, pp.
26-29.
Hardy, B., Ramanantsoa, M.M., Hanneton, S., Lenay,
C., Gapenne, O. and Marque, C. (2000). Cognitive
processes involved in the utilisation of a simple
visuo-tactile sensory prothesis, ISAC'00, Exeter,
England, pp. 52-55.
Kaczmarek, K.A. and Bach y Rita, P. (1995). Tactile
Displays, in Virtual Environments and Advanced
Interface Design, Barfield, W. and Furness III, T.A.
(Eds.), New York : Oxford, 349-414.
Lenay C., Cannu S., Villon P. (1997). Technology and
Perception : the Contribution of Sensory Substitution
Systems. In Second International Conference on
Cognitive Technology, Aizu, Japan, Los Alamitos:
IEEE, pp. 44-53.
Lenay, C., Hanneton, S. & Gapenne, O. (2000). Virtual
space and perception substitution. In Proceedings of
the 4th World Multiconference on Systemics,
Cybernetics and Informatics (SCI 2000), IIIE, Vol.
III (Virtual Engineering and Emergent Computing),
Orlando Floride (USA), pp. 674-677 (selected as the
best paper presented in the session Virtual
Engineering II).
Marque C., Gapenne O., Lenay C.,. Vanhoutte C.,
Stewart J. (2000). The Visual Glove: 2D form
recognition by means of supply technology, in
Proceedings of the Sixth International Conference on
Tactile Aids, Hearing Aids and Cochlear Implants
(ISAC'00), Exeter, Angleterre, pp. 87-90.
O’Regan, J. K. and Noë, A. (2000). A sensorimotor
account of vision and visual consciousness,
Behavioral and Brain Sciences, 24(5), 2000
Paillard J. (Ed.), (1991). Brain and space, Oxford :
Oxford University Press.
Poincaré H. (1905). La valeur de la science, Paris,
Flammarion.
Poincaré H. (1907). La Science et l’hypothèse, Paris,
Flammarion.
Turvey, M. T. and Carello, C. (1995). Some Dynamical
Themes in Perception and Action. In R.F. Port & T.
 
 

  


 
 
! " # %$ $ %
&')( 
*(+(+'

,

-.0/21435.768:9;.0< = >@?3A8:9;.0< =CBD;EGF0HJIK8:HML+NO

-.0EP8RQS35.TU=WVX8 >@Q =WVX8:BD;EGF0HJIK8:HML+NO
YZ\[][_^C`Cacbed+fgahZ\`Ci*j kli'kmd+nebporqd+stZ\ned+fgZ\ngu
v+wxv+wyv*z ah{|d+nga~}Cd+axC€klah{|d w bpoCZ2‚€Z\ned+{A^ w„ƒ ^C`…‡†ˆu\Z\fgZŠ‰Œ‹+ wxŽ\v\ C W;d+‘5d+`

’”“•—– E˜=CD –
™›šœŸž š¡£¢m¤Ÿš—¥p¦ ¡¨§ŸœŸ¦ ¡¨¢‡§ ¦ ©;šª;¢‡«;š—ž ¬£­m¡£¢;§¯®&š—°g©;œŸ¢;¡£¥ ®&¥7±Ÿ²
³x´¶µP·¹¸Gº‚·¹»‡¼¾½xº¹º¿·¿À|½Á»|µP½ ¦ ©;œŸ¦Âᣬ¨¬4œŸ¬¨¬£±e ž ±ŸÄűŸ¦ ¥¦ ±&°—±Ÿ®&®ª;Æ
¢;¡£°—œŸ¦ šÇÂᣦ¾©*©mª‡®&œÈ¢*ÄŚɡ¨¢‡§È¥œŸ¢;«*Ê;œŸž ¦ ¡£°—¡£Ê;œŸ¦¾š ¡£¢ˆ©mª‡®&œÈ¢
¥p±Ë°É¡¨œŸ¬ œŸ°—¦ ¡¨¤!¡¨¦ ¡£š—¥—Ì_Íc¢Î¦ ©;¡£¥ÏÊ;œŸÊŚɞ—ÐXÂњˆ«;šÉ¥ °—žp¡¨ÄRš*œŸ¢Š±Ÿ¢;Æ
¦ ±Ÿ§ÈšÉ¢;š—¦ ¡£°¯®&±m«;š—¬Ñ±Ÿ²¦ ©;š¯¥p±Ë°É¡¨œŸ¬Ñ¡£¢m¦ š—¬£¬¨¡£§Ÿš—¢;°ÉšÈÐ5Âé;¡£°y©¡£¥
ÄŚɡ¨¢‡§Î¡£®&Ê;¬£š—®&š—¢!¦¾šÉ« ±Ÿ¢ ±Ÿª;ž ¡£¢;²¹œŸ¢m¦*ž ±ŸÄűŸ¦—ÐÒ »¶Óx¸˜»‡´˜·¹Ô Ì
Õ ©;šÖž ±ŸÄűŸ¦7¡£¥7°—ª‡ž ž š—¢m¦ ¬£­×°—œŸÊ;œŸÄ‡¬¨š±Ÿ²:œŸ¦ ¦¾šÉ¢m¦ ¡¨±Ÿ¢;œŸ¬2¡£¢m¦ š—ž Æ
œŸ°—¦ ¡¨±Ÿ¢]Âᣦ¾©]©mª‡®&œÈ¢]°—œŸž š—§Ÿ¡£¤Ÿš—ž ¥ÉØ|Âњ,œŸž šÏ«;šÉ¥ ¡£§Ÿ¢;¡£¢;§ˆ¦ ©;š
ž ±ŸÄűŸ¦—ÙÚ¥:¡£¢;¢;œŸ¦ š°—œŸÊ;œŸÄ‡¡¨¬£¡¨¦ ¡£š—¥:¦ ©;œŸ¦5Âᣬ¨¬š—¢‡œÈ欨š¦ ©;šž ±ŸÄűŸ¦
¦ ± œŸ°—Ûmª;¡£ž šÎ¡£¢m¦ š—¢m¦ ¡£±Ÿ¢;œŸ¬¨¡£¦„­ ÜݧŸ±ÈœŸ¬£ÆÞ«;¡£ž š—°É¦¾šÉ« ÄŚɩ;œe¤m¡¨±Ÿž ¥pß
¦ ©;ž ±Ÿª;§Ÿ© š—àmʇ¬¨±Ÿž œŸ¦ ¡¨±Ÿ¢ ±Ÿ²Ç¦ ©;šÊ;©!­m¥ ¡£°—œŸ¬œŸ¢;« ¥p±Ë°É¡¨œŸ¬š—¢‡Æ
¤m¡£ž ±Ÿ¢;®&š—¢!¦Ð:œŸ¢;« ¦ ±ª;¢‡«;š—ž ¥p¦ œÈ¢‡«š—®&ʇœÈ¦ ©;šÉ¦ ¡¨°—œŸ¬£¬£­Š±Ÿ¦ ©;Æ
š—ž ¥ÉÙmÄŚɩ;œe¤m¡¨±Ÿž…Äm­áœŸ¥ °Éž ¡¨Ä‡¡¨¢‡§¦ ±¦ ©;šâ¡£¢m¦ š—¢m¦ ¡£±Ÿ¢ ¦ ©;œŸ¦ÑÄŚɥ ¦
š—àmʇ¬¨œŸ¡£¢;¥7¦ ©;šÄŚɩ;œe¤m¡¨±Ÿž—Ìâãâ¥7¦ ©;šÊ;ž šÉž š—Ûmª‡¡¨¥p¡¨¦ š—¥7²¹±Ÿžâ¦ ©;¡£¥
š—®&ʇœÈ¦ ©;šÉ¦ ¡¨°ˆœŸ¥ °Éž ¡¨Ê‡¦¾¡£±Ÿ¢WÐXÂњ&œŸ¥ ¥pª;®&š*¦ ©;œŸ¦ä ´G·å»‡¼ ¸G¼Þ¼¾½x»‡æ …a ƒ ^Cngk›‹ 9 ú1;<;ú‚ùþ ‡fgoCkˆ^C‘C‘tkln w fgZ\ngigZ)ah`8¾d+`|f,ngZ\stZ\f

=

¼Þ·¹´G» œŸ¢;«›ç ½¹è ½~阷¹ê˜½ë·¹ìí·¹¼Þ¸G¼Þ·¹´G» Ê;¬£œe­¯¡£®&ÊűŸž ¦ œÈ¢!¦Ãž ±Ÿ¬£š—¥¡£¢

¦ ©;š,«;šÉ¤Ÿš—¬£±Èʇ®&š—¢m¦±Ÿ²X¥p±Ë°É¡¨œŸ¬4°—±Ÿ®&®ª;¢‡¡¨°ÉœÈ¦ ¡£±Ÿ¢WÌ
^Cngk‹?e1,oCa~bpo]ahiÃstklah` ƒ kl[_stk!}C}k!}*ah`×fgoCk o|^C[Îd+`]igZKb'a~d+÷
î\ïÃð4ñCòëóÖôõCð4öÁòï kl`|øAahngZ\`C[]kl`|fd+`} stklah` ƒ @ sCngZ\^ ƒ o‡f^C‘8A ,ahfgo o|^C[Îd+`
bld+ngk ƒ ahø\klng
i
€ZKb'a~d+÷h÷hu”ah`|fgke÷h÷¹a ƒ kl`|fÏngZ\stZ\fgiÏ[_^Ci'fÏod¶ø\k)d+` ù;úWûcùÉüýŸúWýÈû„þÞÿ %oCk ÞZ\÷h÷hB
Z ,ah` ƒ igk!b'fgahZ\`Ci }klilbgngahstk Z\^Cn Z\`|fgZ ƒ ke`Wkefga~b
þ
—ûc
ù  fgod+fahiUiga¹[]ah÷~d+nUfgZ fgod+fZÏo|^C[Îd+` ah`|fgkeng÷¹ZKb'^ w []ZK}kl÷Z &igZKb'a~d+÷ah`|fgke÷h÷¹a ƒ kl`b'kÅC  ,oCa~bpo ahi stklah` ƒ ah[]‘W÷hk w
fgZ\ng
i €odÈngah` ƒ iga¹[]ah÷~d+n Z\`|fgZ ƒ ke`Wkefga~b%sd\bp{ ƒ ngZ\^C`}Ão|^ w []kl`|fgkm} Z\` 9 1ú ;<;ú‚ùþ  %oCk Z\`|fgZ ƒ ke`Wkefga~b []ZK}kl÷*od+i
[Šd+`Wid+`} ngZ\stZ\fgibld+` ^C`};klngigfed+`}  a k ‘Cngk!}a~b'frd+`} fgoCngklkˆigfed ƒ kl
i DEg?‹  d\*b F˜^WahigahfgahZ\`rZ Ñah`|fgke`AfgahZ\`d+÷hahf uRG ,oCa~bpo
b'Z\`|fgngZ\÷CfgZˆigZ[]k&}k ƒ ngklk  k!d\bpo]Z\fgoCkl
n  istklod¶øAahZ\
n KfgoCk kl`d+sC÷hkli fgoCkrngZ\stZ\f)fgZ [Šd+{\k^Ci'krZ b'klngf d+ah` []klfgoCZK}i
ngZ\stZ\fg i âklngk ‘Cngkl‘CngZ ƒ ned+[)[]k!} d\blb'Z\ne}ah` ƒ fgZ*d sC÷h^Wke‘Wngah`‡f ÞZ\nŠZ\sCfed+ah`Cah` ƒ ƒ ZŒd+÷hi!H
  v ×a~}kl`|fg,a +bld+fyahZ\` ,ahfgo Z\fgoCklngim
Zd ÿ—ù ¨ýÈûcý%ngkl‘tklngfyZ\a¹ngk)Z 7o|^C[Îd+` stklod¶øAahZ\ngim2klig‘tk w ,oCa~bporkl`d+sC÷hkli,ahf,fgZ×ah`};ahngk!bgfg÷hu * k )|‘Xklngahkl`b'k&Z\fgoCklng
i ;stk w
b'a~d+÷h÷hu&fgod+ f ÞZ\n ø\klngsd+÷‚d+`}Ö`CZ\` w ø!klngsd+÷4b'Z\[][_^C`Cacbed+fgahZ\`… od¶øAahZ\nm ,oCklngJ k I¾Z\ah`|fd+fgfykl`AfgahZ\` d+`}ng*k KC*k )Aahø\kÖah[]a¹f d+fgahZ\`
ah f âZ\^C÷~} }Z âkl÷h÷Z\`C÷hu ah` fgoCk ‘Cngk!}klfgken'[]ah`Ck!} iga¹fy^d w ‘C÷~d¶u)ah[]‘tZ\ngfed+`|f7ngZ\÷hklimŒd+`} #LÑigZKb'a~d+÷b'Z\[][_^C`Cacbed+fgahZ\`…
fgahZ\`Cim7d+`} igZ â k âZ\^C÷~} fed+{\k ;ý
—þ£üú
—û ;úWÿ—ýÎah`Ci'fyk!d\} ahM ` ,oCa~bpofgoCk ngZ\stZ\f^C`};klngigfed+`}iZ\fgoCklng
i |stklod¶øAahZ\ns|u
Z þÞúWûcýŸúWûcþÞù;ú !
—û ;úWÿ—ý #"ëkl`C`Wkefgfm*‹+  $ ah` fgngu|ah` ƒ fgZ d+ilbgngahsWah` ƒ fgZfgoCkah`|fgke`AfgahZ\` d+`}Ukl[]Z\fgahZ\` fgod+fâstkligf7*k ) w
^C`};klngigfed+`} ahfgiÃstklod¶øAahZ\
n &%oCahiÇ[]k!d+`Cifgod+fÃfgoCk,ngkl‘tkln w ‘C÷~d+a¹`ÎfgoCkˆstklod¶øAahZ\
n
fgZ\ahngk od+i¯fgZstk þÞúWÿ—ù ¨ýÈûcýd+`} 'ù WýÈú fg( Z Þ^CngfgoCkln *k ) w
fgke`WigahZ\`Wi&d+`} []ZK},a +5bed+fgahZ\`Ci î\ïON'P…ïÖòöÁóHQÏðSROT î\ïON'P…ïÃð U òWV ò ð
%oCkngZ\stZ\fgiÇ[_^Ci'f,d+÷hi'Z od¶ø\k*d+` ú ;û -.¨þ
—ûcþÞÿˆý 0/—1ù ;þ 2
UýÈú‚û Rd֑Co|uAiga~bld+÷XstZK}u]fgod+f ahi7igfgng^b'fg^Cned+÷h÷huÎd+`3 } Þ^C`b w X k stk ƒ ah3` ,ahfgofgoCk ‘Cngkl[]a¹igk fgod+fâd+`|u¯igZKb'a~d+÷h÷hu]b'Z\[][_^ w
fgahZ\`d+÷h÷hu iga¹[]ah÷~d+n fgZ o|^C[Îd+` igkl`WigZ\nga w []Z\fgZ\n iguAi'fykl[]i `Ca~bld+fgahø\kah`|fgke÷h÷¹a ƒ kl`b'k[_^Ci'fâod¶ø\k¯d]ú ;û-.¨þ
—ûcþÞÿ*ý0/—ù12
%oCk ngZ\stZ\fgi ah`|fgkened\b'f ,ahfgo fgoCk kl`|øAahngZ\`C[]kl`|f ah` fgoCk þ UýÈúWûe!a k RdíngZ\stZ\fâfgod+f7ahi7igfgng^b'fg^Cned+÷h÷huÎd+`}0Þ^C`b'fgahZ\` w

iedÈ[]! k d¶uÎd+i7o|^C[Îd+`CiÖ}Z2!ah[]‘W÷ha~b'ahfy÷¹u]igo5d+ngah` ƒ fgoCklahnâ*k ) w d+÷h÷hu iga¹[]ah÷~d+n×fgZ o|^C[Îd+` igkl`WigZ\nga w []Z\fgZ\nUiguAi'fykl[]i %oCk
‘tklngahkl`b'4 k ,ahfgo o|^C[Îd+`CimÇd+`} ƒ klfiga¹fy^d+fgk!} ah` fgoCk”kl` w ngZ\stZ\fah`|fgkened\b'fgY i ,ahfgo_ahfgikl`|øAahngZ\`C[]kl`|fÑah`¯fgoCkâiedÈ[]C k d¶u
øAahngZ\`C[]kl`|figo5d+ngk!} ,ahfgo o|^C[Îd+`Ci #5W÷~d+fgklø: ‹+\\ 6 7Öi d+iâo|^C[Îd+`Ci}Z2!ah[]‘W÷ha~b'ahfy÷¹u×igo5d+ngah` ƒ ahfgiâ*k )|‘Xklngahkl`b'Z k ,ahfgo
d ‘tZ\i'igahsC÷¹k kl[_stZK}ah[)kl`‡f ÞZ\n fgoCk igZKb'a~d+÷_ah`|fgke÷h÷¹a ƒ kl`b'kÅ o|^C[Îd+` ah`|fgkeng÷¹ZKb'^CfgZ\ngim:d+`} ƒ klfyi&iga¹fy^d+fgk!}Šah` fgoCkUkl`|øAa w
âkísC^Ca¹÷hfd+`”^C‘C‘tkln w fgZ\ngigZah8 ` ¾d+`|f ngZ\stZ\fm: 9 1ú ;<;ú‚ùþ  >=…a ƒ\w ngZ\`C[)ke`|fÏigo5d+ngk![ } ,ahfgoÎo|^C[Îd+`C\ i #5W÷~d+fgklø:5‹+\\ 6
 self (robot) environ-
ment
drives perception
affordance
repertoire
of behaviors
change
value system action

reinforcement

…a ƒ ^CngkyLz ]7_b6FË^Cahi'ahfgahZ\`rZah`|fgke`AfgahZ\`d+÷hahf u
=

ah` d”b'÷h^Cfgfyklngk!}Šsd\bp{ ƒ ngZ\^C`}& v Ö}klfgkeng[)ah`Cah` ƒ ngZ\^ ƒ oW÷hu

…a ƒ ^Cngk v
= ]= ‚ZGø\k!d+fyk!}rd\b'fya¹ø\kˆøAahigahZ\`roCk!d\} ZW9 ú1;<;ú‚ùþ fgoCk }ahngk!b'fgahZ\` ZÏfgoCk o|^C[Îd+` ¾d\b'k”stklah` ƒ fgned\bp{\k!}#L
fgned\bp{Aah` ƒ Z\s I¾k!bgfgc i ,ahfgo iedÈ÷hahkl`‡f_b'Z\÷hZ\n_d+`}›fgk6)Afg^CngkRŒk ƒ
fgZGuAim{  >gz,‘tZ\ah`|fgah` ƒ fgZÎZ\nˆngk!d\bpoCah` ƒ Z\^Cf|ÞZ\nˆfgoCk]Z\s I¾k!bgf
Z\p n ¾d\b'k]s|u ^Ci'ah` ƒ fgoCkÎd+ng[]i)d+`}fgZ\ngigZ2{   a  ƒ ?d _lah` ƒ d+÷ w
^Ï^CnŠngZ\stZ\fmZ9 ú1;<;ú‚ùþ ëigoWZ1,` ah` =…a ƒ ^Cngk ‹\ ahiÎstk w fgkeng`5d+fykl÷hu stklf âklkl` fgoC} k ¾d\b'kd+`} Z\s I¾k!bgfmd+`}  ‰ ø\Z w
ah` ƒ b'Z\`Ci'fyng^b'fgk!} d+i d ‘tZ\i'igahsC÷¹k `d+fg^Cned+÷hahi'fyacb›kl[_stZK}a w bld+÷h,a _lah` ƒ bld+`CZ\`Ca~bld+÷…sd+sWsC÷hah` ƒ ,ahfgo ÷hah‘ w iguA`b oCah` ƒ *Y^Cn w
[]kl`|f3ÞZ\n_fgoCk b'Z\[][_^C`Cacbed+fgahø\k }klø\kl÷hZ\‘C[)ke`|f  †ˆZ_lah[Šd ngkl`|fg÷huR~ âk*d+ngZ k âZ\ng{Aah` ƒ Z\` []ZK}^C÷¹ke i ÞZ\n ƒ ?d _lkfgned\bp{Aah` ƒ 
` 5W÷cd+fyklø  v\Ž\Ž\Ž 6 09 ú1;<;ú‚ùþ ‘tZ\i'igkligi'kli×d+‘C‘WngZ1)Aah[Šd+fgkl÷hu
ah[]‘tklnn Þkmb'f ø\klngsd+÷ah[]a¹f d+fgahZ\`…ëd+`} igZ Z\`…Ãah` Z\ne}kln fgZ
fgoCkÏiedÈ[]k{Aah`Ckl[Šd+fga~bÇigfgng^b'fg^CngkZ 2fgoCkÏ^C‘C‘tklnÇstZK}uZ Ñd ‘CngZGøAa~}[ k 9 1ú ;<;ú‚ùþ l,ahfgofgoCk¯sd+i'a~b‘Co|uAiga~bld+÷ig{Aah÷h÷hiZ  ‰ w
fgoCngklk w u\k!d+n w Z\÷~}ro|^C[Îd+` ah8` ¾d+`|
f UY^Cngngkl`|fg÷huR v a }k ƒ ngklkli fgZ w  w [)Z\`‡fgo w Z\÷~};imíd+i d+` ah`Cahfga~d+÷áigfed ƒ k ÞZ\n›igZKb'a~d+÷d+`}
ZbÞngklkm};Z\[ #"c^d=…6i Je $ ah` fgoCk_oCk!d\}f Lah` fgoCk_`Ck!bp{  b'Z\[][_^C`Cacbed+fgahø\k]}klø\kl÷hZ\‘C[)ke`|f

‰ah` k!d\bpo d+ng[  6k )Kb'÷h^}a¹` ƒ fgoCkod+`58}   d+`} Lah` fgoCk
fgng^C`C[ { e d+ngk_d+ngned+` ƒ k!}ah` c d g \Ž+w [)[ w f d+÷h÷t^C‘C‘tkln7stZK} u € õ‚ëôÖöƒAöcð4öÁòï òWN î\ïÃðST ïÃð4öÁòÇï!P&„Áöcð~…
9 1 ú ;<;ú‚ùþ  ahi[]Z\^W`‡fgkm} Z\` d”fed+sC÷h k ÞZ\H n ¾d\b'k w fyZ w @d\b'k ah` w
fgkened\b'fgahZ\h ` ,ahfgo d*o|^C[Îd+` bld+ngk ƒ ahø\kln,iga¹fyfgah` ƒ ah` d)bpod+ah
n  ` +d `|u†ÞZ\ng[ ÷hah{\kUd‡âZ\ne})Z\n ƒ kligfg^CngkRCdˆiguA[_stZ\÷ah`Š÷~d+` w


ƒ ^d ƒ k ahi`CZ\f*d¯÷~d+stkl÷:ZÑngk*Þkengkl`Afm|sC^Cf*d¯‘Cahk!b'k*Z…oCa ƒ o w
9 1 ú ;<;ú‚ùþ  od+i d ÞZGø\k!d+fyk!} igfgklngklZ øAahigahZ\` oCk!d\})d+i ‘tZ\fgke`Afga~d+÷&ah8` ÞZ\ng[Îd+fyahZ\` ÞngZ\[ ,oCa~bpo fgoCk ngk!b'kea¹ø\kln }k w
igoW1Z ,` ah` =…a ƒ ^Cngk v 3iÏd\bpo Z  fgoCk)klu\kli od+i¯f âZ b'Z\÷hZ\n ngahø\klifgoCk,igkl`5}kl
n  iah`|fgke`AfgahZ\` fgZˆ[Šd+`W,a  kli'fÃigZ[]klfyoWah` ƒ ah`
Y, Y " bld+[]klned+i ÷hah{\k fgoCZ\igk Z  jÃùÉü >7¯}Cd+[)imUklf d+÷ h fgoCkkl`|øAahngZ\`C[]kl`|ˆ f  Z\nâah`fgoCk[]kl[)Z\ngSu igZd+iâfgZ bpod+` ƒ k
v\Ž\Ž\Ž 6k]fgoCk ÷hBZ âkln Z\`Ck od+i d ,a~}k d+` ƒ ÷hk ÷hkl`Ci fgod+f fgoCk ngk!b'kea¹ø\kl
n  i d+fgfykl`AfgahZ\` d+`} ‰+Z\n stklod¶øAahZ\ned+÷¯}ahi'‘KZ\i'a w
ig‘5d+`Ci fgoCk øAahig^dÈZ ÷ +Ckl÷~}  d+stZ\^Cf ‹ v\Ž }k ƒ ngklkli oCZ\ng,a _lZ\` w fgahZ\`  €‘tklngstkln ` X ah÷higZ\`Nj+  ‰ 6 }%oCahi*kl`d+sC÷hkli ^Ci*fgZ
fed+÷h÷hSu e:d+`}ÎfgoCkU^C‘C‘tkln&Z\`CkUod+i×d×fgke÷hkl‘WoCZ\fyZŠ÷hkl`Ciˆfgod+f ‘Cngk!}a~b'f&d+`}rb'Z\`|fgngZ\÷ Z\fgoCklng
i ‡stklod¶øAahZ\E n ÞZ\n,*k ŠŠb'ahkl`|fˆb'Z w
fed+{\kli”d b'÷hZ\igk w ^C‘ ah[Šd ƒ k›Z\` fgoCl k ÞZGø\k!d  d+stZ\^Cf v\Ž }k w Z\‘tklned+fya¹Z\` d+`} b'Z\[]‘tklfyahfgahZ\` ,ahfgo Z\fgoCklng
i |%oCkngZ\stZ\fm
ƒ ngklkliÎoCZ\ng,a _lZ\`|fed+÷h÷hSu 6 %oCngklk []Z\fgZ\ngi }ngahø\k fgoCk klu\klim fgoCklng6k ÞZ\ngkRGod+ifgZ&÷hk!d+ng`]fgoCk,^Ci'k Z tiguA[_stZ\÷hifgZUb'Z\[][_^ w
b'Z\`|fgngZ\÷h÷hah` ƒ fgoCklahn}ahngk!b'fgahZ\`  ‘d+` d+`} b'Z\[][)Z\` fgah÷hmf 6 `Ca~bld+fgkah`|fgke`AfgahZ\`C
i z BZ  bld+`fgoCkngZ\stZ\f}klngahø\kah`|fed+` ƒ a w
%oCk []Z\fgZ\ngi d+÷hi'Z oCkl÷h‘Ci”fgoCk klu\kli”fgZ ‘tklnn ÞZ\ng[ d iedb w sC÷hk*ah`|fgke`AfgahZ\`CE i ÞngZ\[ Z\sCi'klngø‡d+sC÷hkˆstklod¶øAahZ\nÏZ Z\fgoCklng6i ‹
bld\}kZ ZGø\klo n g a }k ƒ ngklklc i ,ahfgoCah` ‹ Ž\Ž []i'k!bm5d+o i âkl÷h÷âd+i %oCk%ngZ\stZ\t f ,ah÷h÷Çstkd+sC÷hk%fgZ^C`};klngigfed+`} Z\fgoCklng
i tah` w
ig[)Z|Z\fgo ‘C^Cn'ig^CahfˆZ âøAahig^dÈ÷fed+n ƒ kefgi [%oCk)ah[Šd ƒ kep i ÞngZ\[ fgke`AfgahZ\`_s|uˆ[]k!d+`CiZ Ca~}kl`|fg,a +bld+fyahZ\Œ ` ,ahfgo¯fgoCkâah`|fgkeng÷¹ZKb'^ w
fgoCk bld+[]klned+i d+ngk Þk!} ah`|fgZ [Šd+i'iga¹ømkl÷hu ‘d+ned+÷h÷¹ke÷&ah[Šd ƒ k fgZ\
n %XZ }Z igZ…fgoCkÎngZ\stZ\f]od+i)fgZ stk d+` þÞúWûcýŸúWûcþÞù;ú 
‘CngZKb'kligi'Z\ngi >rq47ts uÖahigahZ\ ` cÞZ\3n ¾d\b'a~d+÷Öd+`} `CZ\` w ¾d\bgacd+÷ /—ýÈþÞú‡ü bld+‘d+sC÷hk Z  ƒ ZŒd+÷ w }ahngk!b'fyk!}ig‘tZ\`|fed+`CklZ\^Ci7stklod¶øAahZ\n
Þk!d+fy^Cngk fgned\bp{Aah` ƒ B  ,oCa~bpo]kl`d+sC÷hkliÃngk!d+÷ w fgah[]kád+fgfykl`AfgahZ\`d+÷ s|urahfgigkl,÷ < v ` Z\ne}klnÖfgZd\*b F˜^Wahngk_ah`|fgke`AfgahZ\`d+÷hahf uRdngZ\stZ\f
ah`|fgkened\b'fgahZ\` ,ahfgo fgoCkah`|fgkeng÷¹ZKb'^CfgZ\nd+`} ,ahfgo d fgoCahne} igoWZ\^C÷~} ‘tZ\i'igkligiÏfgoC| k ÞZ\÷h÷hBZ ,ah` ƒ y
i  †ˆZ _lah[ŠdŒ v\Ž\Ž ?‹ 6D
Z\s I¾k!bg
f v ` d\}C};ahfyahZ\`RfgoCk_oCk!d\}”od+iÖ÷hah‘CO i ,ahfgo v "o^E=…i
,oCa~bpo d+÷h÷hBZ  fgoCk []Z\^Wfyo fgZ Z\‘tkl` d+`} ig[)ah÷hk ÞZ\ n ¾d w  7
—ýȎú
—ù ¶þ 2Uùû„ù 
r

û„ý   ,ahfgo ,oCa~bpo fgoCk
b'a~d+÷Ã*k )|‘Wngkligiga¹Z\`Ci]d+`}÷hah‘ w iguA`b oCah` ƒ ,ahfgo ø\ZKbld+÷h,a _!d+fya¹Z\w ` ngZ\stZ\fÖbld+`›^Cfgah÷h,a _lkáfgoCk?d tZ\ne}Cd+`b'káfgod+f kl[]kln ƒ kli
iÏd\bpo "o^E= ahi b'Z\`|fgngZ\÷h÷hk!} s|u ah`|fgkeneb'Z\`C`Wk!bgfgk!} q W Y xÖ
i k stklf âklkl`ÎfgoCkˆngZ\stZ\f&d+`} fgoCkˆkl`|øAahngZ\`C[]kl`|
f
oCa ƒ o w ÷hklø!kl÷Xigkl`WigZ\nga w []Z\fgZ\n7ah8` ÞZ\ng[Îd+fyahZ\`”ahi ‘CngZKb'kligi'km}Us|u  7 Áý Wý ¶û„ùþ ÁýŠ'ù ;\/—ý ‘ÈþÞù 
‚  ,oCZ\i'kah`Cahfga~d+÷b'Z\` w
d]b'÷h^Ci'fyklnÏZ q ah`|8^ ) sÇY
i fgke`Afgid+ngkah`C`5d+fykng*k KC*k )Akei!W k ƒ ƒ ned+i'‘Cah` ƒ ,od+f w
9 1 ú ;<;ú‚ùþ  od+i stklkl` k F˜^Wah‘C‘tk!} ,ahfgo fgoCk ÞZ\÷h÷hBZ ,ah` ƒ klø\kln,fgoCkˆod+`5})fgZ\^bpoCkli
Þ^C`b'fgahZ\`C
i D g?‹ _fgned\bp{Aah` ƒ d+` `CZ\`Wig‘Kk!b',a +b o|^C[Îd+` ¾d\b'k  7 igklf×Z 0¶þ ‘È1ý
ÉÑ÷hah{\kŠo|^C` ƒ ken×Z\y n ¾d+fya ƒ ^WkRfgod+f
 ¨«£r£§
’“”•—–>˜z“'™
š ’“'›
œ*ª 
œ*Ÿž1 ¢¡m£¤ ªrœ¤Ä£rËÆÍ̟£¤ žÈ¢É>£ªÊ  ªrœ¤Ä£rËÆÍ̟£¤ žÈ¢É>£ªÊ 
¬ œ*ª ¢­ ¬ ¨«£r£§®­

¤ÇžÈž  ¤ÇžÈž 
¥¦£§©¨
¬¾ ­¿ªrœ*ÀŸ ÂÁäģoÅÃÆ®¤Ç£1ª ¢Æ©ž ¬Î ­bƮş£m ¢Æ©¨«Ïd ¢œ¤ÄË1£r 

…a ƒ ^Cngk|gf ]iÇ[]‘5d+fyo|u4ÞZ\n&d+`CZ\fgoCkln
 iÏstklod¶øAahZ\n

=
…a ƒ ^Cngk
= a ÎYngk!d+fyah` ƒ ¾I Z\ah`|fád+fgfykl`AfgahZ\`l,ahfgo d×bld+ngk ƒ ahø\kln

fgnga ƒ\ƒ klngiÇZ”Z\`CkZ\nÏd*b'Z\[_sCah`5d+fyahZ\`%ZXstklod¶øAahZ\ngi fgoCkb'Z\[]‘C^Wf d+fgahZ\`d+÷âb'Z\igfÖZÃigkl÷hk!b'fgah` ƒ d+`}›igk ƒ [)ke`|fgah` ƒ

ah`rfgoCkˆngkl‘tklngfyZ\a¹ngk. ‘tZ\i'igahsC÷¹kˆb'÷h^CkliCÞngZ\[ fgoCkø|d+i'fkl`|øAahngZ\`C[]kl`|fd+`}%d+÷hi'Z ah`
 7 ‘¯-ÅO ý
r

û„ý  fgod+f…klø|d+÷h^d+fgkli&,od+ffgoCk ngZ\stZ\f [Šd+{Aah` ƒ fgoCk×b'Z\[][_^C`Cacbed+fgahø\k*ah`|fgkened\b'fgahZ\` b'Z\oCklngkl`|
f
‘tklneb'kea¹ø\kli  stZ\fgo *k )|fgklngZtb'kl‘Cfya¹Z\` d+`} ‘CngZ\‘WngahZ w =…a ƒ ^Cngk a ah÷h÷h^Wigfgned+fykli oCBZ  fgoCk ngZ\stZ\f b'ngk!d+fykli d+`}
b'kl‘Cfya¹Z\‚`  °  ÞZ\nŠah`Ci'f d+`5b'k ‘C÷hk!d+i'^Cngk d+`} }ahi'‘C÷hk!d w [Šd+a¹`‡fed+ah`CC i I¾Z\ah`|f&d+fgfykl`AfgahZ\ ` ,ahfgo d]bld+ngk ƒ ahø\kl
n D
ig^Wng k
‹ E%oCk ngZ\stZ\f bld+‘Cfg^Cngkli fgoCk }ahngk!b'fgahZ\` Z  fgoCk

¨r
ý ¶úWþÞú‡ü Uý—ÿ ;úWþ
r fgod+f ngklah8` ÞZ\neb'kl i  ‘tZ\i'a w
 7
bld+ngk ƒ ahø\kl
n  i)d+fgfykl`AfgahZ\` s|u bld+‘Cfg^Cngah` ƒ fgoCkr}ahngk!b w
fgahø\kl÷huÎZ\nÖ`Ck ƒ d+fgahø\kl÷h±u dstklod¶øAahZ\n_d\blb'Z\ne}ah` ƒ fgZ fgahZ\`Z …fgoCk]bld+ngk ƒ ahø\kl
n  istZK}uR4d+ng[][ i  ngkmd\bÁoCah` ƒ ‰
fgoCkˆø|d+÷h^CkˆZ fgoCkˆngklig^W÷h~f ‘tZ\ah`|fgah` ƒ eG ¾d\b'kÅ‚d+`} ‰+Z\n ƒ ?d _l k
² k ƒ a¹`C`Cah` ƒ ,ahfgo ah`C`5d+fyk ng*k KC*k )Akei!W
 ,oCa~bpo b'Z\`Ci'ahigf%Z Ud v E%oCkÏngZ\stZ\f }ZAkli,digk!d+nebpo ah`)fgod+f,}ahngk!b'fgahZ\`Šd+`}
b'Z\`|fgah`|^CZ\^Wi ig‘tk!b'fyng^C[ Z\` igkl`WigZ\nga w []Z\fgZ\n []ZK}Cd+÷ha¹fyahkli! a~}kl`|fg,a +CkliÎfgoCk Z\s I¾k!bgfÎZ ¯fgoCk bld+ngk ƒ ahø\kln d+fgfykl` w
fgoCkëngZ\stZ\f*k )|‘W÷hZ\ngklifgoCk ƒ d+[_^CfZ 4*k tkmb'fgahøŒ k  ‘CngZ +Cfed+sC÷hk  fgahZ\`w
ÿ -.
—ý 2 ý ³ëýÿÈûÑd+i'igZ4b'a~d+fgahZ\`CifgoCngZ\^ ƒ oahfgiah`|fgkened\b'fgahZ\Œ
r ` ,ahfgo
Lz E%oCkngZ\stZ\f ZKblbld+igahZ\`5d+÷h÷hu }ahø\klngfyiahfgid+fgfykl`AfgahZ\`
fgoCk kl`|øAahngZ\`C[]kl`|fmˆd+iah÷h÷h^Wigfgned+fyk!} ah` =…a ƒ ^Cngk Lz %oCk
ngZ\stZ\f ahi ƒ ned\}^5d+÷h÷hu d+sC÷hk fgZ ^Ci'k fgoCkligk d+i'igZ4b'a~d+fgahZ\`Ci sd\bp{rfgZ fgoCkbld+ngk ƒ ahø\klnÖfgZbpoCk!bp{r,a ÃoCkZ\nÖigoWk_ahi
ig‘tZ\`|fed+`CklZ\^Cig÷¹u%d+O i UýÈû 1ù 2Þü;ù td+i'igZ4b'a~d+fgahZ\`C
i X k od¶ø\k igfgah÷hÐ
÷ ÞZKb'^Ci'ah` ƒ Z\`ÎfgoCkˆZ\s I¾k!bg
f
}*k +C`Wkm} fgoCahiˆd+iÏfgoCk×d\*b F˜^WahigahfgahZ\`ÎZ ah`|fgke`AfgahZ\`d+÷hahf  u €fgnga~b'fy÷hu ig‘tk!d+{|a¹` ƒ Ñ I¾Z\ah`|f d+fgfykl`AfgahZ\` ngk F˜^WahngkliU`CZ\fZ\`C÷hu
  d ÒÞZKb'^Ci'ah` ƒ Z\`)fgoCkÏiedÈ[]kÏZ\s I¾k!bgfm˜sC^CfÏd+÷hi'Ó Z ‚s  [_^Cfg^d+÷
î\óOT ïÃð4ö´Öõ‚P5ð4öÁòï µ ö~ðSR ¶ ð±R!T ñ8ƒ d\bp{A`CBZ ,÷hk!} ƒ kl[]kl`|fÏZ fgoCahiÏigo5d+ngah` ƒ d\b'fya¹Z\w `
7ÖiâigoW1 Z ,`”ah ` =…a ƒ ^Cngk ‰Œ: 9 1ú ;<;ú‚ùþ Šb'ngk!d+fyklid+`}[Šd+a¹` w
%XZ ^C`};klngigfed+`} Z\fgoCklng
i áah`|fgke`AfgahZ\`Cim)fgoCk ah`|fgke`AfgahZ\`d+÷ fed+ah`Ci I¾Z\ah`|f d+fgfykl`AfgahZ\` ,ahfgo fgoCk o|^C[Îd+` bld+ngk ƒ ahø\kl
n
ngZ\stZ\f,od+ifgZ]kl[]‘d+fgoCklfga~bld+÷h÷hu a~}kl`|fga, u›ahfgigkl÷,Y,ahfgorZ\fgo w =…a¹ngigfmÑahfgiבtklngah‘CoCklned+÷ w ø|a¹6 k  bld+[]klned+i]igk!d+nebpo ÞZ\n›dro|^ w
klng4 i #"íd+^Wfykl`Co5d+oC`…‹+\ $ 6k4fgoCahi&ngkF˜^WahngkliáahfáfgZ›ah[Šd ƒ ah`Ck [Šd+` ¾d\b'kah` d”b'÷h^Cfgfyklngk!}”øAa~}klZ”ieb'kl`C k o^Ï`b'k 3 d ¾d\b'kahi
oCBZ  Z\fgoCklng} i Þklke÷ˆd+`} d\b'f >=…a ƒ ^Cng( k gz6 7Öird`|^C[_stkln }klfgkmb'fgk!};fgoCk&klu\kli iedbebld\}kˆfgZ×fgoC‡ k ¾d\b'kUd+`})ni ,a¹f bpoÎfgZ
Zˆngkligk!d+nebpoCklngi  .k ƒ ² d+ngZ\` w YZ\oCkl`*‹+\ a k&€Cbld+i'ke÷h÷cd+fyaÁ fgoCŒ k ÞZGø\k!d+÷ w øAah*k  bld+[]klned+c i ÞZ\nd”b'÷hZ\igk w ^C‘ ah[Šd ƒ kZ ÃfgoCk
v\Ž\Ž\Ž k4%XZ\[Šd+i'kl÷h÷hZ2 ‹+\\ Šig^ ƒ\ƒ kligfgkm}5J  ·+ù;þÞúWlû ;ûcûcýŸúWûcþÞù;ú ¾d\b'. k ‡=‚ngZ\[ fgoCahi&ah[Šd ƒ kÅKahf&ngZ\^ ƒ oW÷hu kligfgah[Šd+fgkli&fgoCk›}a w
‘C÷~d¶uAi]d+` ah[]‘tZ\ngfed+`|f*ngZ\÷hk]ah` ^C`};klngigfed+`}ah` ƒ oCBZ  Z\fgo w ngk!b'fyahZ\` Z âfgoC\ k ¾d\b'y k ÞngZ\[ fgoCk]ig‘5d+fya~d+÷ëd+ngned+` ƒ kl[)ke`|f*Z 
klngH i Þklkep
÷  6k )AfgkengZXb'kl‘Cfya¹ø\kl÷h±u m \xÖ`}klngigfed+`}a¹` ƒ oCBZ  Z\fgo w fgoCd k ¾d\b'a~d+÷ b'Z\[]‘tZ\`Ckl`|fgid+`} d+÷hi'ˆ Z ÞngZ\[ fgoCkÏZ\‘Cfga~bld+Ž÷ KCBZ 
klngid\b'f ngk F˜^Wahngkl° i Áý ¸ rý ¹;þ ‘Ÿý þ UþÞû ;ûcþÞù;pú ÞZ\n:fgned+`Cig÷~dÈfyah` ƒ fgoCk Z\`ŠfgoCo k ¾d\b'.k ¿%oCkl`…  9 1ú ;<;ú‚ùþ igfed+ngfgiëigk!d+nebpoCah` ƒ ah`Šfgod+f
Z\sCi'klngø!k!}]d\b'fya¹Z\`UZ KZ\fgoCklngiah`|fgZˆfgoCkëngZ\stZ\
f  iBZ ,`][]Z\fgZ\n }ahngk!b'fgahZ\` d+`})a~}kl`|fg,a +CkliÏfgoCkˆZ\s I¾k!bg f ,ahfgoÎiedÈ÷hahkl`‡fˆb'Z\÷hZ\n
‘CngZ ƒ ned+[ Z\nϑCngZ\‘WngahZ4b'kl‘Cfya¹Z\` d+fgf d\bÁoCk!})fgZ]fgod+f&d\b'fya¹Z\w ` d+`} fg6k )Afg^Cng k
Ô Z\ng[Šd+÷ ah8 ` ¾d+`|fgi”d+`} bpoCah÷~};ngkl` d\bpoCahklø\y k I¾Z\ah`|f d+fgfykl` w
28:.0V –Š’ –l– 14V – .J8 V
º
fgahZ\` *k tZ\ngfy÷hkli'ig÷h u kmfgoCklu”d+÷hi'Z ƒ ^Ca~};kZ\fgoCklng
i |d+fgfykl`AfgahZ\` s|u
‘tZ\ah`|fgah` ƒ d+`} ƒ ?d _lah` ƒ d+f fgoCklahnÃd+fgfykl`AfgahZ\`d+÷fed+n ƒ ke
f ŸZ\ah`|f
ŸZ\ah`|f&d+fgfykl`AfgahZ\`Šahi fgoCkUd\b'f Zigo5d+ngah` ƒ k!d\bpoÎZ\fgoCkln
 iád+f w d+fgfykl`AfgahZ\`)ahiZ\sCi'klngø!k!}¯ah`rdø\klnguk!d+ng÷huigfed ƒ kÏZ Xah8` ¾d+`|fg
i 
fgke`AfgahZ\`d+÷ÞZKb'^Ci » f ig‘tZ\fg÷ha ƒ o|fgiZ\s I¾k!bgfgiÇd+`}klø\kl`|fgiE d+`} }klø\kl÷hZ\‘C[)ke`|
f k ahf igfed+ngfgt i Þ^C`b'fgahZ\`Cah` ƒ st*k ÞZ\ngkΉ[]Z\`|fgoCi
b'Z\`b'kl‘Cfyi!Èah`×÷~d+fgkln…igfed ƒ kliZ4ah`8¾d+`|fgi
\}klø\kl÷hZ\‘C[)ke`|f65stk w d+`} stk!b'Z\[]k []Z\ngk igZ‘CoCahi'fya~bld+fgk!} ^C‘ fgZ ‹  []Z\`|fgoCi
ah` ƒ d+fgfykl`}k!}fgZ_s|u)f âZZ\nâ[]Z\ngk֑d+ngfga~b'ah‘d+`|fgi7Z b'Z\[ w  ² ^Cfgfgken n  Z\ngfgo ` ;d+ngngklfgfËR‹+\Œ?‹ 6 »7,ffgoCC k +Cngi'f igfed ƒ kReah` w
[_^C`Cacbed+fgahZ\`…Wfgo|^Cib'ngk!d+fyah` ƒ dUigo5d+ngk!} b'Z\`|fg6k )Afíah` ÞngZ\`|f ¾d+`|fgi bld+`×a~}kl`|fg,a  ufgoCkád+fgfykl`AfgahZ\`d+÷Œfed+n ƒ kefÃah`×fgoCk ngZ\^ ƒ o
ZWfgoCkl[ ¼%oCk7igo5d+ngk!}b'Z\`|fg6k )AfÑahi7d ig^WsCigklfÑZ WfgoC k âZ\ng÷~} }ahngk!b'fgahZ\`  .k ƒ |nga ƒ o|fZ\n÷h*k Þfigac}k  Z …fgoCk)d ƒ kl`|
f  ioCk!d\}
fgoCkUb'Z\`Ci'fyahfg^Ckl`|fgiÏZ ½,oCa~bpo d+ngk&[_^Cfg^d+÷h÷hu d\blb'kei'iga¹sC÷hk&fgZ Z\`C÷hu ,oCkl` fgoCkrah8` ¾d+`|fgi igklkstZ\fgo fgoCk d ƒ kl`|fÎd+`} fgoCk
fgoCkŠ‘d+ngfga~b'ah‘d+`|fg
i kÇahf]‘C÷~d¶uAiŠdr[Šd I¾Z\n]ngZ\÷hkÎah` ngk!}^b'ah` ƒ fed+n ƒ kefd+f7fgoCkÖiedÈ[]kfgah[] k Ò7Ö÷higc Z I¾Z\ah`|fd+fgfykl`AfgahZ\` d+f7fgoCahi
 …a ƒ ^Cngk
= $ ]= d\b'a~d+÷ k*)|‘Wngkligiga¹Z\`CiÏZW9 ú1;<;ú‚ùþ

…a ƒ ^Cngk׉z 9 1ú ;<;ú‚ùþ ‡kl` ƒ d ƒ ah` ƒ ha `}I¾Z\ah`|fˆd+fgfykl`AfgahZ\`

=
` q”ZAZ\ngkR,‹+\\  d+`} Uþ Áù úW1ý -.pù;ú
 j ,a _*_lZ\÷~d+fyfga `
,ahfgo d*o|^C[Îd+` lb d+ngk ƒ ahø\kln
7ÖngsCahs…‚‹+\  6 Ô keZ\`5d+f d+÷C[]ah[)a~b {Aah` ƒ ZKigZ[]k¾d\b'a~d+÷Ck*) w
‘Cngkli'igahZ\`Ciˆahi!4oCZBâklø\klnmKigZŠngkligfgnga~b'fyk!}rfgod+fáahfU}ZAkli&`CZ\f
Þ^C÷h÷hu d\blb'Z\^C`|] f ÞZ\nfgoCk }klø\kl÷hZ\‘C[)ke`|f Z á[]Z\ngk igZ‘CoCahi w
igfed ƒ k ahiZÞfykl`÷hk!}%s|uŠfgoCk]bld+ngk ƒ ahø\kln
 ingk!d\}ah` ƒ ah`8¾d+`|fgi
 fga~bedÈfyk!}ah[]a¹f d+fgahZ\` fgod+fˆah`|ø\Z\÷hø\klˆ i ,oCZ\÷hk)stZK} u ]q”ahngngZ\n
d+fgfykl`AfgahZ\`d+÷4}ahngk!b'fgahZ\`w ¼%oCahiÑng^};ah[]kl`‡fed+ngu*f uA‘tk ZI¾Z\ah`|f `Ckl^CngZ\`WZ i ÞZ\^C`} ah`fgoCk¯‘Cngk w [)Z\fyZ\n¯b'Z\ngfy*k )ŠZ [Šd\bld FË^Ckli
d+fgfykl`AfgahZ\` ahiëfgoCkíZ\`Ckífgod+d f 9 1ú ;<;ú‚ùþ  ahiíb'^Cngngkl`|fg÷hurbld+‘d w d\b'fya¹ø|d+fgk ,oCkl` fgoCklu Z\sCi'klngø!k igZ[]klZ\`Ck }Z\ah` ƒ d ‘d+n w
sC÷hk*Z < fga~bg^W÷~d+n d\b'fya¹Z\` d+`} ,oCkl` fgoCklu }Z fgoCk iedÈ[]k d\b'fya¹Z\`
fgoCkl[]igkl÷hø\kl
i z BZ âklø\klnmËfgoCk*b'÷~d+ah[ fgod+fÇ[]ahn'ngZ\n `Ckl^CngZ\`Wi
d+ngkfgoCk_ah`C`5d+fyksd+i'ahO i ÞZ\nÖah[]a¹f d+fgahZ\` ahiÖ`CZ\f_b'÷hk!d+nmŒiga¹`b'k
1:Ö1|?.Ä×t1MØË< . – = – .J8:V
Õ
[Šd\bld FË^Ckliˆ}Z]`CZ\fÏah[]a¹f d+fgkUd+f&d+÷h° ÷ #%XZ\[Šd+i'kl÷h÷hZ25‹+\\ 6
j k*KCk*)Aahø\kíah[]a¹f d+fgahZ\`›ahiád&igklf Z ngk*KCk*)Akeiëfgod+fë‘CngZK}^b'kli %XZ k )|‘W÷~d+ah` fgoCk Z\nga ƒ ah` Z ”ah`};ahngk!bgf *k )|‘Xklngahkl`b'k Z 
*
d b'klngf d+ah` d\b'fya¹Z\` ah` ngklig‘tZ\`Ci'k”fgZ Z\sCi'klngø|ah` ƒ dŠiga¹[]ah÷~d+n Z\fgoCklng
i  stklod¶øAahZ\nmY  âk d+i'ig^C[)k›fgod+fU`CklZ\`d+fgkliבtZ\i'igkligi
d\b'fya¹Z\`ÎZ Z\fgoCklng
i X k×d+ngkˆklig‘tk!b'a~d+÷h÷hu ah`|fgkengkli'fyk!}%ah`ÎfgoCk U1 ù } Z\n iguA`Ckli'fyoCklfga~*b U‘tklneb'ke‘WfyahZ\`  ² d+ngZ\` w YZ\oCkl`…
ng*k KC*k )Akei…fgod+fÑigoW,a ÞfZ\`C k  i7d+fgfykl`AfgahZ\M `  k. ƒ s|u Z\ngahkl`|fed+fyahZ\` ‹+\\‰ e:ah` ,oCa~bpo stZ\fgo *k )|fgklngZtb'kl‘Cfya¹Z\`  Z ,øAahig^dÈ÷Á fed\b w
Z fgoCkÖoCk!d\8} …ah` ngklig‘tZ\`Ci'kÖfgZZ\sCi'klngø|ah` ƒ igZ[]klZ\`CkÖkl÷hig k  i fgah÷hkRAklf b àÎd+`} ‘CngZ\‘WngahZ4b'kl‘Cfya¹Z\`  Z …ah`C`Wke! n Þklke÷ha¹` ƒ i‘CngZ w
d\b'fya¹Z\`  ig^bpo d+i_oCk!d\} ngZ\fed+fyahZ\` d+`} d+ng[ igfgngklf bÁoCah` ƒ   }^b'k!( } ÞngZ\[ stZK}u ‘tZ\i'fy^Cngkli]d+`}[]ZGø\kl[]kl`|fg6i  d+‘C‘tk!d+n
stk!bld+^Cigk%fgoCklu }ahngk!b'fg÷hu ah8` KW^Ckl`b'k Z\`C k  i d+fgfykl`AfgahZ\` d+`} ah`Šd,iga¹` ƒ ÷hk ig‘5d\b'k ig‘5d+`C`Ck!}ˆs|u}ah[)kl`CigahZ\`WiÃig^5b oŠd+iÃig‘5d w
stklod¶øAahZ\ned+÷}ahi'‘KZ\i'ahfyahZ\& `  8` ¾d+`|fgi d+`} d\}^W÷hfyi›od¶ø\k d fga~d+Þ÷ ‰+fykl[)‘tZ\ned+{ ÷ Þngk F˜^Ckl`b'uŒ…d+[]‘W÷hahfy^5}kR…d+`} k ƒ ZKb'kl`|fgnga~b
fgke`5}kl`b'ufgZ ah`|ø\Z\÷h^C`|fed+ngah÷hu×igoW,a ÞffgoCklahnÏd+fgfykl`AfgahZ\`)fgZ*fgoCk ÷hZKbld+÷h,a _!d+fya¹Z\` ÷hah{\k ÷h*k Þf w iga~}k Z\n”nga ƒ o|f w igac}k Z *fgoCk igkl,÷ ~ i
}ahngk!b'fgahZ\` Z t¾d\b'a~d+÷d+`} `CZ\` w ¾d\bgacd+÷íigfgah[_^C÷h4 a > f d+{A^CnedŒ stZK} u %oCahi d+[]ZK}Cd+÷֑tklneb'ke‘WfyahZ\` âZ\^C÷~} ‘CngZK}^b'k ngk w
v\Ž\Ž ‹?6 KC* k )Aahø\këah[]a¹f d+fgahZ\`…Ÿ÷hah{\këfgod+fZ S¾d\b'a~d+÷Œ*k )|‘Wngkligiga¹Z\`CimŸoCk!d\}
%oCahiÏ{Aah`}  Z ng*k KC*k )Akei*b'Z\`C`Ck!b'f&},a tklngkl`|f,[]ZK}Cd+÷ha¹fyahkli k ngZ\fed+fyahZ\` d+`}ΑCngZ\fgZ w ngk!d\bpoCah` ƒ ² k ƒ a¹`C`Cah` ƒ ,ahfgo FË^Cahfgk
Z\`CkŠù /
—1ý ‘Èý
d+`CZ\fgoCkln‘tklngigZ\& `  istZK}u›*k )|fgklngZtb'kl‘Cfya¹ø\kl÷hu d×ngZ\^ ƒ o [Šd+‘W‘Cah` ƒ ‚fgoCkUng*k KC*k )Aahø\kUah[]a¹f d+fgahZ\` âZ\^C÷~} ƒ klf
 [Šd+ah`C÷hu øAahig^dÈ÷h÷hS u &d+`} Uù ‘Èý
 d\bgfg^d+÷h÷hu Z\n&øAahngfy^5d+÷h÷h±u  +C`Ck w fy^C`Wk!}fgoCngZ\^ ƒ oigZKb'a~d+÷ ah`|fgkened\b'fgahZ\` k ƒ Aah[]a¹f d+fgahZ\`
 .

Z\`C k  i¯BZ ,` stZK} u %XZ ƒ klfyoWket n ,ahfgM o I¾Z\ah`|f)d+fgfykl`AfgahZ\`…2ngk w ‘C÷~d¶±u W,ahfgo bld+ngk ƒ ahø\klng
i
KC* k )Aahø\k›ah[]a¹f d+fgahZ\` kl`d+sC÷hkliUfgoCk›ngZ\stZ\ffgZ bld+‘Cfg^Cngk”oCBZ   ` d\}C};ahfyahZ\` ng*k KC*k )Aahø\k ah[]a¹f d+fgahZ\` Z E¾d\b'a~d+÷ ƒ kligfg^Cngkli
igZ[]klZ\`Ck kl÷higk ;ý—ý ¯
d+`} ;ÿÈû
 %oCahi ahi þÞú ;þ Áý—ÿÈû”rý ¹2 oCkl÷h‘Ci_ah8` ¾d+`|fgi%d+`} bld+ngk ƒ ahø\klngi_fgZigo5d+ngk%kl[]Z\fgahZ\`d+÷Ïd+i w
W ý ¶þÞýŸúWÿýR:ah` ,oCa~bpo fgoCkngZ\stZ\f_ah`};ahngk!bgfg÷hu *k )|‘Xklngahkl`b'kli ‘tk!b'fyi Z  fgoCk ah`|fgkened\b'fgahZ\`  ² ngk!?d _lk!d+÷ ` €Cbld+i'igkl÷h÷~d+fgaÁ
igZ[]klZ\`CkÇkl÷hig k  istklod¶øAahZ\J n ÙÚ klklÜ÷ Û¶d\b''f Ý2s|u&fgned+`Cig÷~dÈfyah` ƒ ah`|fgZ v\Ž\Ž\Ž 6 Р[]ahfed+fgahZ\` Z оd\b'a~d+÷t*k )|‘Wngkligiga¹Z\`Cim!÷hah{\kÖfgoCZ\igk_bld+`
fgoCk_ngZ\stZ\
f  iBZ ,`  d\bgfg^d+÷Z\nøAahngfy^5d+Þ÷ 7stklod¶øAahZ\y n Ùr—Þklke÷ ¶ stk‘CngZK}^b'k!}s|t u 9 1ú ;<;ú‚ùþ d+i igoW1Z ,` ahˆ ` =…a ƒ ^Cngk $ m âZ\^C÷~}
 d\bg6 f ÝÈWd+i,‘CngkløAahZ\^Ci'÷hu igoW1Z ,`ahh ` =…a ƒ ^Cng| k gf ah`};^b'këiga¹[]ah÷~d+nkl[]Z\fgahZ\`×ah`×fgoCk ah8` ¾d+`|fg
i &%XZ ƒ klfyoWkeb n ,ahfgo
7 `|^C[_stkln7Z 5ngkligk!d+nebpoCklngiâod¶ø\kÖig^ ƒ\ƒ kligfgkm}fgod+f ‘tklZ w I¾Z\ah`|fd+fgfykl`AfgahZ\`…Ëah8 ` ¾d+`|fgid+`}bld+ngk ƒ ahø\klng i âZ\^C÷~}stk d+sC÷hk
‘C÷hk d+ngkah`C`5d+fykl÷hu k F˜^Wah‘C‘tk!} ,ahfgo fgoCk d+sCah÷¹ahf¾u fgZ bld+‘ w fgZÖigo5d+ngk kl[]Z\fgahZ\`Ci…fgBZ d+ne}ifgoCklah‚ n I¾Z\ah`|fg÷hu)d+fgfykl`}k!}áfed+n w
fg^Cngk d+`CZ\fgoCkln ‘tklngigZ\& `  i d\b'fya¹Z\`C
i k igZ[]k Z  fgoCk klø|a w ƒ klfyi k!fgoCahv i âZ\^C÷~}Z Þfykl`stkZ\sCi'klngø!k!}ah` fgoCZ k ÞZ\ng[ Z 
—ù 2
}kl`b'kliífgoCklu b'ahfyk!} d+ngkrúWý—ù;ú ;û YUþ þÞ1ÿ ßÈþÞú;ü #q”kl÷hÂf _lZ  ÿÈþ &Áý ;1ý pýÈú‚ÿÈþÞú;üÑ ,oCklngk×ah8` ¾d+`|fgiˆ÷hZAZ\{ah`|fgZÎfgoCklahnUbld+ngk w
 ² d+ngZ\` w YZ\oCkl`р‚ Cg‹+\ a  ì þÞú /£þÞú ;úWý1
r
<í]îáú éE
r
<
ù;úÓîo-Rûcþ
r ;ú  ï ý—ù)ù';°ì þÞú  Žq4r% sngkligi
² d+ngZ\` w YZ\oCkl`ր‚ og‹+\\‰  iŠfgoCklngk d `CZ\ng[Îd+÷á‘CodÈigk
indirect
experience
Z  iguA`d+kli'fyoWkei'a~dah` }klø\kl÷hZ\‘C[)ke`|6f ‹‡wðE
rËÿ ýŒuÃZ\÷ v 

Ô f Z v $
² ngkm? d _lk!d+÷ÁW‡ Y ` €Cbld+i'igkl÷h÷~d+fgaÁ ² ‚ v\Ž\Ž\Ž   8` ¾d+`|f w ÷hah{\kˆigZ w
i, o i’, o’
b'a~d+÷*ah`|fgkened\b'fgahZ\`Cistklf âklkl` d ngZ\stZ\f d+`} d o|^C[Îd+`
projection estimation bld+ngklf d+{klnmz îZ'5ûcþ ‘Èo ý ñÖý ‘ÈþÞù RG uÃZ\÷   Ô fZ h5‹
of intention of intention ² ^Cfyfgkln n 7Z\ngfgo…d  ò\ ` d+`} ;d+ngngklfgfË Ô !g‹+\Œ?‹  X od+f
…a ƒ C^ ngk  7Öieb'ngah‘CfgahZ\`rZ d+`CZ\fgoCkln
 iÏstklod¶øAahZ\n,fgZ]fgoCk
= []ah`5}irod¶ø\k ah` b'Z\[][)Z\` ahirig‘5d\b'.k D ig‘5d+fya~d+÷ˆ[]k!bpo w
ah`|fgke`AfgahZ\` d+`})kl[]Z\fgahZ\`rfgod+f,stkligf,k*)|‘W÷~d+ah`CiÏahf
d+`Cahi'[]i”igklngøAah` ƒ I¾Z\ah`|f”øAahig^dÈ÷¯d+fgfykl`AfgahZ\` ah` ah8` ¾d+`b'uR
ñO¶þÞûc þ
ó ù -.¶Žú  'ù ; ô_1ý ‘È1ý ¨'ù ýŸúWû  ðE
rËÿ ù £ùɏü \
uÃZ\ ÷ WŒ‡‘Cw ‘ aa?ëz$\v 5‹+\Œ‹
² u|ng`WkÅmp j Gg‹+\ a  ï y ý ï þÞú ßÈþÞú‡E ü îW| ý õ éE‘Èù ö-Åû„þÞùŽú 
ƒ ha ø\klngi
8¾d\b'k0,oCkl` fgoCkluod¶ø\kUkl`b'Z\^C`|fgkengk!}ÎigZ[]klfyoWah` ƒ ÷è¶þ£ü;þÞú
á' ù ;°9 úWû„ý >¨þ£ü;ýÈúWÿýR» ^E):ÞZ\ne[ } xÖ`Cahø\klngigahf u sngkligi
,oCZ\i'kˆø|d+÷h^CkÓ k. ƒ ;ied kˆZ\nˆ}Cd+` ƒ kengZ\^Wi6ahiÏ^C`b'÷hk!d+n

"ád+^Cfgke`Wod+oC`…RÓ † ½g‹+\ $  áb'Z\^C÷~} stk¯u\Z\^ e fgoCk¯‘CoCk w
`CZ\[)ke`WZ\÷hZ ƒ a~bld+÷Ñ}ah[)kl`CigahZ\`ÎZ  igZKb'a~d+÷2^C`};klngigfed+`}ah` ƒ 
á :öÁï!â
T ã òWä ä ôÖïöÁõ‚P5ð4öå{T j¿~/É ý ¶ú‚ýÈûcþÞ1ÿ
‡;ú lê 

—ûcý 0
ó ù -.¶Žú p. uÃZ\÷ v\  Ô fZ a 
%oCk d+sCah÷¹ahf¾u fgZ a~}kl`|fga, u ,ahfgo Z\fgoCklngi d+÷h÷hZB,i fgoCk ‘Cw ‘ Gg2‹ $?ë g a Lz
" kl`C`CklfyfmG  "3 ‡ Y 8g‹+  $  ï è ý 9 úWû„ýÈúWûcþÞù;Žú ŽêWøû ú‚ÿýRù q4r%
ngZ\stZ\ffgZ ^C`};klngigfed+`} kl[]‘d+fgoCklfga~bld+÷h÷hu Z\fgoCklngi
 ah`|fgke` w
sngklig i
fgahZ\`Cid+`} ‰+Z\nÇkl[]Z\fgahZ\`Ci stkloCah`}_fgoCklahn øAahigahsW÷hkstklod¶øAahZ\n

 fed+{|^CnedŒë‚ € E v\Ž\Ž ?‹  j *k KC*k )Aahø\krigoW,a Þf)Z _d+fgfykl`AfgahZ\`  ah`
%oCk7ngZ\stZ\f d+ilbgngahstkliÑfgoCkâah`};ahngk!bgfg÷hu** k )|‘Xklngahkl`b'k!}Östklod¶ø w
ahZ\náfgZ›fgoCk[]kl`|fed+÷igfed+fgkÅ‚÷hah{\kUah`|fgke`AfgahZ\` d+`}Îkl[]Z\fgahZ\`… ;d+‘5d+`Ckligk e ó ' ;ú êWùŸÿÈþÞýŸû  'ù ;Œô_1ý ‘È1ý ¨'ù ýŸúWû &ðE
r~2
kligfgah[Šd+fgk!} s|u ^Ci'ah` ƒ
—ý æ;<2¦pý ¸ ý—ÿÈû„þ ù;w ú  ` fgkeng[)i Z &ahfgi ÿ ù £ù—ü hõ ê 
Wù
—þ -. ù;Ó ú ú;ú;ü -.Ÿüo ý ô_1ý ‘È1ý ¨'ù ýŸúWû
û ïR ù ß-.
þ 0ü ó ' ;1ú ý8
ZB,` ah`|fgke`AfgahZ\`d+÷hahf uRŒigkl,÷  w n'6k KWkmb'fgahZ\`fgke÷h÷¹ifgoCk¯ngZ\stZ\ffgoCk
[]kl`|fed+÷ igfed+fgkŠfgod+f×stkligf›}klilbgngahstkli)fgoCkŠstklod¶øAahZ\
n kÃfgoCk †ˆZ _lah[ŠdŒ z ‚ v\Ž\Ž ?‹  7Ö` ^Ï`|fgZ ƒ ke`Wkeu”Z €ZKb'a~d+÷h÷huYZ\[ w
ngZ\stZ\ffgoCkl`‘CngZ I¾k!bgfgifgoCahi[]kl`|fed+÷ igfed+fgk sd\bp{ŠZ\`|fgZ%fgoCk [_^C`Cacbed+fgahø\k j Z\stZ\fgim0  9 úWû„ý Ê;ÿÈûcþ ‘Ÿþ
—û ê -.0Uý  9 ú
—ûcþ 2
Z\nga ƒ ah`5d+÷stklod¶øAahZ\] n >=…a ƒ ^Cngk  6 O%oCahiÖahiíoCBZ  fgoCk_ngZ\stZ\f û -Åûcý û 9mê9m2þÿÿ
üÒñÖýÈû £ý ý 0üÐð‚îˆü¼êGîýz


^C`};klngigfed+`}iÏZ\fgoCklng
i ‡[]kl`|fed+÷ igfed+fgkei †ˆZ _lah[ŠdŒ z ` 5W÷cd+fyklø  l v\Ž\Ž\Ž  7Ö` kl‘Ca ƒ kl`Cklfga~b
%oCahirkl[]‘d+fgoCklfga~br^C`};klngigfed+`}ah` ƒ Z _Z\fgoCklngiÎahir`CZ\f

d+‘C‘WngZŒd\b o fgZo|^C[Îd+` w ngZ\sKZ\f%b'Z\[][_^C`Cacbed+fgahZ\`…¼  9 úWû„ý 2
Z\`C÷hu fgoCk×{\klu fgZŠigZKb'a~d+÷ëb'Z\[][_^C`Cacbed+fgahZ\`…tsC^CfUd+÷hi'ZŠfgoCk ú ;ûcþÞù;Žú ù ß
'ù  ù;ú Ïù /—ù;3 û ;ú  çˆ- 3;ú 9 úWû„ý 2
;ÿÈûc þ ‘Ÿý jÃù 30-Åú‚þÞrÿ ;ûcþÞù;ú û H÷¿ì(î




 

{\klu fgZ þ UþÞû ;ûcþ ‘Èý ¨rý ¶úWþÞú‡ü Z ˆigZKb'ahZ w b'^C÷hfg^Cned+÷¯b'Z\`|ø\kl` w
fgahZ\`Ci*ah`b'÷h^5}ah` ƒ ÷~d+` ƒ ^5d ƒ k ˆ []ahfed+fgahZ\` ah\ i FË^d+÷hahfed+fyahøkl÷hu  "! #"%$"&%
},a tklngkl`|Ò f ÞngZ\[ kl[_^C÷~d+fgahZ\w ` fiÇ[_^C÷~d+fgahZ\` ahi fgoCkngkl‘CngZK}^b w ')(+*-,/.1032 54 76 ')080%9/(  ' ;:<>=/& ?%?%?%@ A (+9/B/0%CDB)E CGF
fgahZ\` Z fgoCk›iedÈ[]k›ngklig^W÷hfUs|u []k!d+`Ci×Z *dΑCngk w *k )Aahigfgah` ƒ 9/(  9/(+B/(+CH,1E ,JI-0%C7KMLONHPQI-CRE CH,>I-STIU,VE ,/I-0%C<I-B>I-S  0%9/,1E CH,
stklod¶øAahZ\ned+÷&ngkl‘tklngfyZ\a¹ngk Z\nrZ\`C k  iÎBZ ,` fgnga~d+÷ w d+`} w klngngZ\
n k RW0%9X,/N(+0%9/I-(+BY0%R;NHZS[E C F3(+\%(+*-0  S(1CH, ^] C _ a` E%F3(+* 6
b c Z,J,/(+9/LX0%9/,/N = (aF
B @ edVfhgWij
ijgWk
lhgWlmdVleno
lpqr ? #
ah[]a¹f d+fgahZ\` b'Z\‘Cahkli¯fgoCk%ah`|fgke`AfgahZ\`d+÷Ç^Ci'k%Z  []klfgoCZK}t i ÞZ\n
Z\sCfed+ah`Cah` ƒ ƒ ZŒd+÷h] i  ² uAng`CkRы+\ a 6 O%oCkb'Z\[]‘tklfykl`b'Œ k ÞZ\n %$ s E Sut9/IvF
w%(mxyCI-\%(+9/B/I-,zP A 9/(+B/B 
ah[]a¹f d+fgahZ\`rahiÏig‘tk!b',a +bfg4 Z ç ù UÓ ù
r'5þÞýÈú
d+`} od+i ƒ ahø\kl` {|I-.+.+0%*vE ,/,JI }b y6 4 9/tI-t  '  4 
=/& ?%?%~%@  E Cw%ZGE w%(
fgoCkUig‘tk!b'ahkliáfgoCk d+sCah÷¹ahf¾u fgZ”igo5d+ngkUah`};ahø|a~};^d+÷7b'ngk!d+fyahZ\`Wi LOI-,/NI-C€0%Z9w%91E B  ƒ‚…„‡†lˆ
YgWl}
†e‰Š„‹k
pgW†lpo†…Œ 0%* Ž"& 
d+`}áfgZ[Šd+a¹`‡fed+ah`UfgoCkl[ ZGø\kln ƒ kl`Cklned+fya¹Z\`Cim\b'ngk!d+fyah` ƒ ÷~d+` w ;& ~%~ #& ? 
‘’ E B“(1*-*jE ,JI  c”Q=•Ž%–%–%–%@ — N(+0%9/P 0%R˜STI-C™F RW0%9 E NHZš
ƒ ^d ƒ k]d+`}rb'^C÷hfg^Cngkˆah`rfgoCkˆ‘CngZKb'kligy i #%XZ\[Šd+i'kl÷h÷hZ25‹+\\ 6
^Ï^CnZ\` w„ƒ Z\ah` ƒ d+fgfykl[]‘Wf fgZ Z ÞZ\igfgklb
n 9 1ú ;<;ú‚ùþ &ah`¯fgoCk igZKb'a~d+÷ S˜E CD0%IvF 9/0%tr0%, dJ›œ››ƒžDmŸD  dVlDi¡†„¢lD
ijgWk
lD
£¥¤¦k
leno†„¢
†lpo†hk
l<§
‰…f€
lk3gWˆ¨©k
ªk
ijgWp 
kl`|øAahngZ\`C[]kl`|Y f ,ah÷h÷|fgke÷h÷‡^Ci[]Z\ngkd+stZ\^CffgoCk ‘Cngklngk FË^Cahiga¹fykli
‘  (+9/tr(+9 ¬« r6 ­ I-*-B/0%C r« ƒ=/& ?%~%!%@ ©†£®†¯e3lp“†°)¤¦k
f€
xú ; û -.Áý fÞZ\n d+`}íZ\`|fgZ ƒ ke`Wkefga~b‘CngZKb'kligZ i xú -.¶û -.Áý 2Z CfgoCk
fh‰ŠlDgWp
ijgWk
l 
lDˆ ¤¦k±3lgWijgWk
l²[³ E 9/\HE 91F xyCI-\%(+9/B/I-,zP
d+ngfg,a +b'a~d+÷‚igZKb'a~d+÷Wstklah` ƒ iÇfgod+f,bld+`rb'Z\`|fgngahsC^CfgkfgZ*fgoCkÏo|^ w
[Šd+` b'^C÷hfg^CngkˆfgZ ƒ kefgoCklE n ,ahfgoÎ^Ci A 9/(+B/B 
— S˜E B“(+*-*-0  ' =/& ?%?%?%@ ‚M´
†µ¤¦‰Š£¶i¡‰Š„‹
£X„¢g®±
gWlDukn§
‰…
%
0
fh
l ¤¦k±3lgWijgWk
l·³ E 9/\HE 91FTxyCI-\%(+9/B/I-,zP A 9/(+B/B 
U TÐN'T ñ8T ïÖõfT҃ ¸D*jE ,J(+\  _ 5=/& ?%?%?%@ — N(¹(  I-w%(+C(+B“I-By0%R¦ST(^E CIUCwºI-C NHZš
_}Cd+[]i! ² h ² gn kmd?_lk!d+÷ÁY‡ h ² ngZAZ\{Aim2jp h ` €Cbld+i'igkl÷h÷~d+fgaÁ
7
S˜E Cºtr(+I-Cw%B  E CGF  0%B“B/I-t*UPhI-Cº9/0%tr0%,/B H»‰Šlˆ ¼lgW¯†„¢
² | v\Ž\Ž\Ž  z ^C[Šd+`CZ\ac} j Z\stZ\fgi
D 7 Ô k6 †ˆah`5} Z gWijq½¤¦k±3lgWijgW¯†€ŸDi¡‰Šˆ3gW†e€¾e¿À  ZCGFÁxyCI-\%(+9/B/I-,zP 
%XZAZ\÷Áè 9Âé{éWé 9 Wú û„ý>¨þ£ü;ýÈúWûyê 

—ûcý0
WuÃZ\÷ ‹ a  Ô Zf ¼g2
‘Cw ‘ v a?ë LŒ‹
 A robotics framework for studying
the coevolution of signaling
Andres Perez-Uribe (Andres.PerezUribe@unifr.ch)1
Department of Informatics, University of Fribourg,
Chemin du Musee 3, 1700-Fribourg, Switzerland
Michele Courant
Department of Informatics, University of Fribourg,
Chemin du Musee 3, 1700-Fribourg, Switzerland
Abstract
In this paper, we propose a robotics framework for
studying the coevolution of signaling. Our motiva-
tion is twofold. First, we propose a situated and em-
bodied framework for signaler-receiver interaction,
and second, we provide a promising approach for
the study of mechanisms that would enable adap-
tive systems to access new information channels
and to exploit implicit information in their envi-
ronments. We present experimental results on a
successful coevolution of signals that enable a very
simple communication between two robots. Finally,
we delineate some aspects of forthcoming research.

Introduction
Communication appears near ubiquitous throughout Figure 1: Signaler-Receiver experimental setup.
nature. Animals (and plants) use signals or displays
to convey information to other organisms (Maynard-
Smith and Harper, 1995). Traditionally, biologists models do not deal with abstract descriptions, but
have investigated the function of animal signals e.g. with a noisy and changing environment; moreover,
by experimenting with articial stimuli. More re- robots have bodies and experience the environment
cently, the use of game theory models has been con- directly but imperfectly, and their actions with the
sidered an important change in the elds of ani-
mal behavior and behavioral ecology (Dugatkin and
Reeve, 1998). Evolutionary games have been in-
spired from economic decision theory describing the
potential interactions of two or more individuals
whose interests do not entirely coincide (Maynard-
Smith, 1982; Dugatkin and Reeve, 1998). Biologists
environment are part of a dynamic system (Brooks,
1991) (See Pfeifer and Scheier (1998) for a complete
reference to embodied cognitive science). Second,
by gaining new insights into how biological systems
evolve strategies to access new information channels
and exploit implicit information in their environ-
ments (e.g., how do spiders Agelenopsis aperta get
have used them to model signaler-receiver interac- to know their opponents' size in a web ght -and
tions to study theoretical aspects of a wide range decide whether to ght or to retreat- by sensing the
of animal interactions, including prey-predator in- vibrations of the web (Richiert, 1978)), we might de-
teraction, communication between rivals, potential velop new concepts for the design of sensory process-
mates, ospring and parents, etc. More recently, the ing mechanisms for adaptive systems. We start by
use of articial neural networks and articial evolu-
tionary techniques has provided a useful framework
y discuss the problems of the noisy
for studying further theoretical aspects of the inter-
action between signalers and receivers (Arak and En-
describing the signaler-receiver experimental setup
and then brie
perception of the robot. The following Section de-
lineates experimental results of the coevolutionary
quist, 1995; Bullock, 1997; Ghirlanda and Enquist,
1998).
In this paper, we propose the use of robots to
study the coevolution of signaling systems. Our mo-
tivation is twofold. First, by using robots we pro-
vide an experimental setup where signalers and re-
ceivers are situated and embodied. This means, robot
1
To whom correspondence should be addressed
process, and nally, the last Section presents some
concluding remarks and future work.
Experimental setup
In our experiments, two autonomous robots are
placed in a box with white walls (Figure 1). We have
used one Khepera robot with a linear vision turret,
which contains a 64  1-cell linear image sensor, giv-
ing a linear image of 64 pixels with 256 gray-levels
each. It has automatic light adaptation (auto iris),
and the frame rate we used was 5 Hz. The second
Khepera robot is provided with a MODEROK mod-
ule, a display module with 16  4 LEDs developed
in collaboration with the Ecole d'ingenieurs de Fri-
bourg (EIF).
The two robots are separated by a distance of
about 8 cms (measured between their central axes),
such that the view angle of the robot with the lin-
ear vision system coincides with the width of the
signaler's display (the robots in Figure 1 are shown
separated by a distance longer than that used in the
experiments). In our experiments, for each signaler-
receiver interaction, the signaler has the possibil-
ity to display one of two possible bar-code-like pat-
terns (denoted here by L and R) determined by
its genotype. The receiver perceives the pattern
and uses an articial neural network with synaptic
weights determined by its genotype, to produce its
response (encoded by two binary outputs). A suc-
cessful signaler-receiver interaction is dened as the
interaction where the receiver outputs a '00' pat-
tern in response to a signaler's L signal and '11' in
response to a signaler's R signal. The additional re-
ceiver outputs '01' and '10' might be interpreted as a
receiver response to an unknown signal or to a signal
it is not concerned with.
The genotype of the signaler is a 32-bit string
which encodes 8 real values in the range [0,1) with a
4-bit precision. Each real value determines the prob-
ability with which a 4  4-LED column is turned-on
Figure 2: Example of noisy signal perception. Each plot
on the right shows the activation of the 64 cells of the
receiver's linear vision system when the signaler displays
the column-based pattern on the left.
run needed over 5 hours time (each generation was
executed in about 8 minutes).
The experimental setup was conceived with the
idea of establishing signaler-receiver interactions at
in the MODEROK display. The rst 4 real values
thus encode the probabilistic L pattern while the
following 4 real values encode the probabilistic R
pattern. Probabilistic patterns were used in order
to simulate some noise in the generation of the dis-
plays.
The genotype of the receiver is 5  16  2 bits
long. It encodes 32 real values equally distributed in
the interval (-1,+1) with a 5-bit precision (one bit is
used to encode the sign). The real values correspond
dierent angles and distances from the very begin-
ning, however, due to technical problems with the
positioning of the robots, we have so far explored
the coevolution of signaling between two immobile
robots. The setup used so far is thus not really
\strongly" embodied, but as well as robot motion
can hardly be simulated, so thus noisy perception.
As an example, Figure 2 shows the linear vision im-
ages obtained by the receiver given some signaler
displays. It appears that a turned-o column acti-
to the synaptic weights of a 16-input, 2-sigmoidal- vates up to 80% the linear vision cells (rst exam-
output perceptron without biases. The inputs to ple of Figure 2), and such activation varies not only
the articial neural network are computed as the
normalized mean activation of groups of four neigh-
boring sensor cells of the linear vision sensor of the
receiver robot. The binary outputs of the receiver
are computed by comparing the perceptron outputs
with a threshold of 0.5.
Embodiment and noisy perception
It should be noted that even if the robots do not
move in these experiments and that they are placed
in a box with white walls, there is a lot of variation
in the sensor readings due to inherent noise of the
with the activation of other 4  4-LED columns (sec-
ond and third examples of Figure 2), but even with
the activation of LEDs in rows that the linear vi-
sion \does not", in principle, perceive (compare the
second and fourth examples of Figure 2): in our ex-
perimental setup, it is the lower row of LEDs which
is maximally perceived by the receiver (e.g., a single
LED in the lower row may activate a linear vision
cell up to 100%). Moreover, a pattern with all the
16  4 LEDs turned-on can be perceived as \noisy"
as shown in the last example of Figure 2.
sensors and changes in the environment due to ex- Articial coevolution of signals
ternal lights. For instance, daylight changed quite We have used the experimental setup described
a lot during the experiment given that a complete above in a coevolutionary robotics experiment
 1 1

0.9 0.9

0.8 0.8

0.7 0.7

Fitness (Receiver)
Fitness (Signaler)
0.6 0.6

0.5 0.5
Best Best

0.4 0.4

0.3 0.3

Average
0.2 0.2
Average
0.1 0.1

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Generations Generations

Figure 3: Coevolution of signaling: signaler and receiver's best and average (of the whole population) tness per
generation. Each point of the plots corresponds to the average of 4 runs.

erating two dierent column-based patterns (asso-

L: R: ciated to the signals) given the genotype-encoded
probabilities of turning-on a particular LED column.
For instance, the L and R patterns in Figure 4 have
0.0% 31% 50% 0.0% 94% 0.0% 50% 94%
a probability of 99.87% of being dierent. It appears
Figure 4: Example of coevolved signals. The gray- that nding a set of probabilities that generates two
level indicates the probability of turning-on a given col- (frequently) dierent patterns is not dicult. For
umn (black represents the maximum probability, which example, the set of probabilities L' = f50%, 50%,
is around 94%, white means turned-o). 50%, 50%g, R' = f50%, 50%, 50%, 50%g generates
two dierent patterns with a probability of 93.75%.
However, L' and R' do not produce coherent pat-
(See Nol and Floreano (1999) for a complete refer- terns, that is, the variation of displays produced by
ence of the basic concepts and methodologies). The the signaler given such probabilities is very large.
sizes of both the signaler and the receiver popula- This impedes the receiver to be able to evolve a
tions are 20, and each signaler-receiver interaction means of distinguishing L' from R'.
is repeated 10 times. For each interaction, the indi- The problem of establishing a successful signaler-
vidual being evaluated (tness computation) is given receiver interaction thus needs a signaler's genotype
0.1 points if the signaler-receiver interaction is suc- that determines two (very often) dierent patterns
cessful, and 0 points otherwise (notice that the t- in coherent manner, that is, the probabilities should
ness is zero also if the receiver responds correctly to tend to be as close to 0% or 100% as possible, and be
only one of the L or R displays). Each signaler in- complementary for L and R. Notice, that the proba-
dividual of a given generation is evaluated by mak- bilistic patterns shown in Figure 4 produce coherent
ing it interact with the best receiver individual of displays in two of the four columns with the highest
the latest 10 generations. When less than 10 gen- probability2.
erations have past, one chooses the best receiver of Another interesting aspect of the coevolved pat-
a random past generation. The same procedure is terns is that of multicomponency (Rowe, 1999) (i.e.,
done for a receiver evaluation. Both populations are signals made up of multiple components, but not
then allowed to reproduce. One-point crossover was necessarily in multiple modalities). In patterns like
applied with probability 0.6, and random mutation those of Figure 4, some components might evolve to
was applied to each bit with a constant probability 'amplify' another component. For example, two of-
of 0.05. We used a roulette-wheel selection mecha- ten turned-on columns might have evolved not only
nism and no elitism. to signal something but to enhance the perception
Figure 3 shows the average tness of signalers and of an often turned-o column (See the above discus-
receivers of four runs of 40 generations each. Fig-
ure 4 shows the coevolved signals for the best sig- 2
Interestingly, this pattern appears to be quite sym-
naler of generation 40 of the rst run. The gray- metric. \Symmetry [in nature] may result from the need
level indicates the probability of turning-on a given among organisms to recognize objects irrespective of the
column. Such a signaler obtained a tness of 1.0 by manner in which they are encountered in the outside
interacting with the best receivers of the 10 preced- world" (Enquist and Arak, 1994). However, in our ex-
ing generations, which means it succeeded in all its periments, we have so far detected this kind of pattern
signaler-receiver interactions. once, and furthermore, given that the robots are immo-
bile, this property is not likely to be evolved by that
We have then calculated the probability of gen- reason.
 31% 25% 56% 0.0% 94% 94% 25% 38%
depended on the morphology of the receiver's per-
ception system) and thus its cells are maximally ac-
L: R: tivated by a particular row of LEDs of the display
(the lowest row in this case). Accordingly, the upper
row cells appear to modulate ('amplify' or 'abridge')
94% 0.0% 56% 44% 0.0% 94% 44% 0.0%
the perception of a given pattern in the lower row of
Figure 5: Example of the coevolved signals for the sec- the signaler's display (Figure 6).
ond set of experiments.
Concluding remarks
100

Within the eld of Cognitive Science, the abundant

evidence of spatial and body-based metaphors in
90

human linguistic expressions, and more recently in

80

mathematical reasoning (Lako and Nun~ez, 2000)

Activation of the linear vision cells

70

60

50
has allowed for postulating shared mechanisms for
40
low-level control of embodied action and higher-level
30

cognition (Mataric, 1997). Thus, there is a growing

interest in both the body-related mechanisms under-
20

lying cognition, and the role of the environment in

10

the development of embodied cognition. Similarly,

0
0 10 20 30 40 50 60
Linear vision cells

Figure 6: Example of the in
uence of the upper row
of LEDs of the signaler's display in the recevier's per-
uence of the sense
ception of the pattern '0100' in the lower row of LEDs.
They appear to change the light intensity detected by
the linear vision and consequently the level of activation
of the cells (due to the auto-iris), in particular those cor-
responding to the turned-o columns of the signaler's
patterns.
some biologists concerned with the study of signal-
ing systems argue that it is the in
organs and nervous systems of an organism, and of
the environment, which drives the selection of sig-
nal precursors (Bradbury and Vehrencamp, 2000).
Therefore, a signaling system arises as the result of
a coevolutionary process between senders and re-
ceivers in which both benet from its use (Zahavi
and Zahavi, 1997). In this paper, we have proposed
sion on the noisy perception and Figure 2). Finally, a robotics framework for studying the coevolution of
it should be noted that once coherent parts of the signaling systems. In particular, we show the suc-
evolved displays enable successful signaler-receiver cessful coevolution of signals that enable a very sim-
interactions, the rest of the displays (i.e., columns ple communication between two robots: a signaler
that turn on and o with similar frequencies) can be robot evolves a couple of column-based patterns, and
a source of bias for the development of future signals. a receiver robot evolves an articial neural network
To further explore the multicomponency issue, we that \interprets" those patterns.
performed a second series of experiments, where we Our experimental framework appears to be a
used the same setup as above, except for the fact promising approach for the study of how to pro-
of using a larger signaler genotype: a 64-bit string vide autonomous articial agents with the ability
encoding 16 real values in the range [0,1) with a 4- to access new information channels (e.g., the sense
bit precision determined the probability with which organs and nervous systems of an organism serve
a 4  2-LED cell was turned-on in the MODEROK many functions besides the detection of and reac-
display. When we devised the rst set of experi- tion to any particular signal (Johnstone, 1998)) in
ments we thought the 4  4 column-based patterns order to develop and continuously adapt their com-
will ease the evolution of signaling given that the munication systems. For example, this will enable
receiver used a linear vision system. Nevertheless,
we found that even if the new experiment involves
a larger search space, the possibility to display pat-
y mentioned along this paper
terns other than 4  4 columns permits the coevo-
lution of a signaling system in the same number
of generations and even with a better performance
(Figure 7). This is due in part to the in
uence of
the dierent rows of LEDs of the display in the ac-
tivation of the cells of the linear vision (Figure 6).
Figure 5 shows an example of the new coevolved sig-
nals: the new two-row coevolved patterns have the
property of having probability values of turning-on
a given 4  2-LED cell closer to 0% and 100% in the
lower row. This was expected because the receiver's
vision system is linear (i.e., the coevolved patterns
autonomous articial agents to coordinate their ac-
tivities in open-ended environments (Steels, 1999).
Many aspects brie
can be further studied with this experimental setup;
researchers (Arak and Enquist, 1995; Bradbury and
Vehrencamp, 2000; Ryan, 1998), for instance, have
recently promoted the notion of sensory drive as a
source for new signals, thus we may ask, what are
those receiver bias that enabled the evolution of par-
ticular signals ? If we let the robots move, what
will be the evolved patterns that enable robust com-
munication ? Would those signals tend to be sym-
metric ? How does multicomponency enhance sig-
nal discrimination ? How might it evolve ? What
does the receiver's neural network learn to be able to
 1 1

0.9 0.9

0.8 0.8

0.7 0.7
Best

Fitness (Receiver)
Fitness (Signaler)
0.6 0.6 Best

0.5 0.5

0.4 0.4

Average
0.3 0.3

0.2 0.2 Average

0.1 0.1

0 0
0 5 10 15 20 25 30 35 40 0 5 10 15 20 25 30 35 40
Generations Generations

Figure 7: Coevolution of signaling (second set of experiments): signaler and receiver's best and average (of the whole
population) tness per generation. Each point of the plots corresponds to the average of 4 runs.

dierentiate the signaler's signals ? What is the ef- and Animal Behavior. Oxford University Press,
fect of perception noise and \interpretation" noise in 1998.
the complexity of the receiver's articial neural net- G. Lako and R.E. Nun~ez. Where Mathematics
work? (Lerena, 2000). Future work will deal with comes from: How the embodied mind brings math-
this and more challenging aspects of coevolving sig- ematics into being. Basic Books, 2000.
naling systems. P. Lerena. Sexual preferences: Dimension and com-
plexity. In From Animals to Animats 6, Proceed-
Acknowledgments ings of the Sixth International Conference on Sim-
The author wishes to thank Patricio Lerena for valu- ulation of Adaptive Behavior, pages 395{404. The
able suggestions and very passionate discussions, MIT Press, 2000.
Dario Floreano for his help with the coevolutionary M. Mataric. Studying the role of embodiment in
algorithm, and the anonymous reviewers for their cognition. Cybernetics and Systems, 28(6):457{
careful reading of a draft version of this paper. This 470, 1997.
work has been supported by the Swiss National Sci- J. Maynard-Smith. Evolution and the Theory of
ence Foundation. Games. Cambridge University Press, 1982.
References J. Maynard-Smith and D.G.C. Harper. Animal Sig-
nals: Models and Terminology. Journal of Theo-
A. Arak and M. Enquist. Con ict, receiver bias and retical Biology, 177:305{311, 1995.
the evolution of signal form. Philosophical Trans- S. Nol and D. Floreano. Evolutionary Robotics.
actions of the Royal Society, London B, 349:337{ The Biology, Intelligence, and Technology of Self-
344, 1995. organizing. The MIT Press, Cambridge, Mas-
J. Bradbury and S. Vehrencamp. Economic models sachusetts, 1999.
of animal communication. Animal Behavior, 59: R. Pfeifer and C. Scheier. Understanding In-
259{268, 2000. telligence. The MIT Press, Cambridge, Mas-
R. Brooks. New approaches to robotics. Science, sachusetts, 1998.
253:1227{1232, 1991. S.E. Richiert. Games spiders play: Behavioral vari-
S. Bullock. Evolutionary Simulation Models: On ability in territorial disputes . Behav. Ecol. Socio-
their character, and application to problems con- biol., 3:135{162, 1978.
cerning the evolution of natural signaling systems. C. Rowe. Receiver psychology and the evolution of
PhD thesis, University of Sussex, 1997. multicomponent signals. Animal Behavior, 58(5):
L. Dugatkin and H. Reeve, editors. Game Theory 921{931, 1999.
and Animal Behavior. Oxford University Press, M.J. Ryan. Sexual Selection, Receiver Bias, and the
1998. Evolution of Sex Dierences. Science, 281:1999{
M. Enquist and A. Arak. Symmetry, beauty and 2003, 1998.
evolution. Nature, 372:169{172, 1994. L. Steels. The Talking Heads Experiment: Words
S. Ghirlanda and M. Enquist. Articial neural net- and Meanings. Best of Publishing, Brussels, 1999.
works as models of stimulus control. Animal Be- (Pre-edition of the book in progress).
havior, 56:1383{1389, 1998. A. Zahavi and A. Zahavi. The Handicap Principle:
R. Johnstone. Game Theory and Communication. In A missing piece of Darwin's puzzle. Oxford Uni-
L. Dugatkin and H. Reeve, editor, Game Theory versity Press, 1997.
 Embodied Intentional Dynamics of Bacterial Behaviour
Catholijn M. Jonker (jonker@cs.vu.nl)
Vrije Universiteit Amsterdam, Department of Artificial Intelligence,
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands. URL: http://www.cs.vu.nl/~jonker

Jacky L. Snoep (jls@maties.sun.ac.za)

University of Stellenbosch, Department of Biochemistry,
Private Bag X1, Matieland 7602, Stellenbosch, South Africa.

Jan Treur (treur@cs.vu.nl)

Vrije Universiteit Amsterdam, Department of Artificial Intelligence,
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands. URL: http://www.cs.vu.nl/~treur

Hans V. Westerhoff (hw@bio.vu.nl)

Vrije Universiteit Amsterdam, Department Molecular Cell Physiology,
De Boelelaan 1087, 1081 HV Amsterdam, The Netherlands. URL: http://www.bio.vu.nl/

Wouter C.A. Wijngaards (wouterw@cs.vu.nl)

Vrije Universiteit Amsterdam, Department of Artificial Intelligence,
De Boelelaan 1081a, 1081 HV Amsterdam, The Netherlands. URL: http://www.cs.vu.nl/~wouterw

Abstract 1987); in his view the notions by themselves need not

Existing (embodied) chemical models of bacteria are
rather complex, due to the thousands of interacting
chemical reactions within the cell. To gain a higher level
of understanding, more transparent and, thus more
abstract, models are needed. Intentional models are
sometimes advocated to gain transparent models of
complex dynamical systems. However, a main problem
with intentional models is the symbol grounding
problem: the gap between the symbolic, discrete binary
decision processes and the continuous flow of chemical
reactions in physical reality. In this paper an intentional
modelling approach based on continuous time is
introduced and used to simulate the behaviour of E. Coli.
The model is grounded in physical reality by precisely
defined relationships to the chemistry. The
intentionalisation approach followed here is relevant in
general, not only for biochemistry. In general, our
approach enables the incorporation of real continuous
time in intentional (BDI) models.
1. Introduction
It is often claimed that the attribution of intentional
notions, such as beliefs, desires and intentions, eases
the understanding of complex agent behaviour; cf.
(Dennett, 1987). However, the relationship of such
attributed intentional notions and the real world is not
uncontroversial. In principle, to relate them to the real
world, two strategies are possible. The first strategy is
to relate the intentional notions directly to physical
circumstances. This strategy is rejected by (Dennett,
be based on any physical substance, as long as they
effectively explain and predict behaviour. The second
strategy is to relate the notions to observed behaviour.
For example, (Dennett, 1991) suggests that the
intentional notions relate to observed behaviour
patterns; however, no indication is given about how
exactly these relationships are defined. This lack of
grounding makes the position of intentional notions to
describe behaviour debatable from a foundational
perspective. From a pragmatic perspective, however,
intentional notions might well have their value in
explanation and prediction of complex behaviour at an
abstract level.
Using symbolic models for intentional behaviour
introduces the well-known symbol grounding problem.
For example, in (Sun, 2000) this problem is discussed,
and an approach is proposed, where a combination of
symbolic and other, e.g., connectionist techniques is
used: the symbols get their grounding by relating them
to lower level (e.g., sensory) processes within the
organism. Also in (Clark, 1997, 1999) a position is put
forward to integrate functional and embodied
perspectives in explanation of behaviour. The work
presented below has a similar perspective: integration
of intentional and embodied models to describe
bacterial behaviour.
For simple organisms such as the bacterium E. Coli,
the chemical processes are sufficiently accessible to
obtain an explanation of, e.g., their eating (food import)
behaviour (Neidhardt, Curtiss III, Ingraham, Lin,
Brooks Low, Magasanik, Reznikoff, Riley, Schaechter
& Umbarger, 1996). However, although biologists in
principle can describe bacterial behaviour by hundreds
to thousands of differential equations for the various
chemical reactions, they want more abstract ways of
summarising the main paths of processes involved. This
motivates for reconsidering the use of intentional
notions for this purpose as well, but to avoid the
foundational problems this time by using these notions
in a grounded, embodied manner. This poses the
interesting question how to relate a discrete, binary
decision process to the continuous dynamics of
(chemical) processes in the real world.
Section 2 briefly describes the bacterial regulation
process, introduces the intentional notions used, and
relates them. In Section 3 the use of temporal
relationships to model both chemical and intentional
dynamics is explained. Subsequently, in Section 4, the
food import decision process of the common bacterium
E.Coli is described using temporal relationships.
2. Relating Chemical and
Intentional Notions in E. Coli
Bacteria are small autonomous living systems that
interact with their environment; the understanding of
the regulation of the behaviour is often complicated by
the enormous complexity of the chemistry in the living
cell. Using intentional notions to model the regulation
of a bacterium, this regulation may be more easily
understood. First, the regulation in bacteria is briefly
explained in biochemical terms. Second, the behaviour
of an agent is explained using intentional notions.
Third, the relationships between the intentional notions
and the chemicals in the bacterial regulation are
presented.
2.1. Bacterial Regulation
In bacteria, as in every living cell, the regulation of its
internal processes consists of several steps (Neidhardt,
et.al., 1996). In this paper, the regulation of the lactose
import is taken as an example; other regulation paths
follow similar steps as depicted on the left side of
Figure 1. First a substance in the cell relates to the
presence of lactose. The transcriptional regulation,
translational regulation and metabolic control then
interact to modify the behaviour of the cell.
So, the regulation of the processes within a bacterium
consists of several steps. First, circumstances in the
external environment lead to certain concentrations of
specific internal substances. Then, the transcriptional
regulation is done, possibly resulting in mRNA.
Subsequently, the translational regulation is done,
possibly resulting in proteins. The metabolism,
comprised of energy production, transport and growth
pathways, further regulates the activation and inhibition
(inactivation) of certain enzymes. When all this is done,
enzymes may be ready to catalyse chemical reactions.
When enzymes catalyse reactions, they cause an
increased flux, leading to growth of the bacterium.
2.2. Intentional Notions
The intentional notions that are used to describe
behaviour are taken from BDI (Beliefs, Desires and
Intentions) models; e.g., (Rao & Georgeff, 1991). The
beliefs represent what the agent deems to be true in its
environment. A belief is present due to sensing (in the
present or in the past). Desires are interpreted as what
the agent wants to accomplish or fulfil. Agents can have
desires contradictory in their fulfilment, for example
desiring lots of ice creams and a slim waist. A desire,
together with a sufficient additional reason, leads to an
intention to fulfil the desire. An additional reason is a
set of beliefs that have to hold or not hold, in order for
the intention to be generated. Intentions are interpreted
as that the agent will make something happen (action),
as soon as a belief in an opportunity (for the action)
occurs. Opportunities are states of the environment that
give the possibility to perform an action. Actions
performed by the agent affect its internal or external
physical environment. The relations between the
intentional notions are depicted on the right side of
Figure 1.
2.3. Intentionalisation
The intentional notions used to describe the behaviour
can be related to the substances used in the bacterial
regulation. The internal substances relating to the
situation in the environment are chosen to correspond
with the beliefs. DNA parts are chosen to correspond to
desires. The conditions needed for the transcriptional
regulation correspond with the additional reasons of the
intentional model. As can be seen from the left side of
Figure 1, DNA is used to create mRNA. Therefore,
with DNA as desire, mRNA is chosen to correspond
with an intention to prepare for an action. The enzymes
created by the translation are used to increase the flux
of chemical reactions (which correspond to actions in
the intentional model). Thus, the enzymes are chosen to
correspond with intentions to perform actions. The
(co)factors necessary for the translation of mRNA into
enzymes correspond with the additional reasons for the
creation of the intentions for performing actions. The
opportunity for an action corresponds to the inhibitors
of the enzymes. When enzymes cause flux, (i.e.,
successfully catalyse reactions), this corresponds to the
action happening in the world.
 Bacterial
Regulation
DNA
activation protein/repressor
mRNA
indicator
environment (co)factor
substances
enzymes
inhibitor
flux
Figure 1: The correspondence between the bacterial regulation and the causation in the BDI model.
3. An Abstract Continuous Time Model
of Chemical Processes
The bacterial behaviour results from a multitude of
biochemical processes. These operate on each other
over time, producing the behavioural regulation. The
overall regulation process is not easy to understand; for
example, a number of feedback loops between different
stages of the regulation process and a high number of
chemical reactions are involved.
A more abstract model for the dynamics of
biochemical processes can be obtained by introducing
categories of concentrations of substances, and relating
different categories of the same and of different
substances over time. In this section temporal
relationships are used to express the timing dynamics.
The resulting abstract model captures the timing
dynamics of the biochemical reactions in logical
temporal relationships using continuous time.
A generally accepted way to describe biochemical
reactions is in the form: A + B ↔ C + D. This expresses
that substances A and B can be transformed into
substances C and D, and that the reverse process is also
possible. In the cell the pathways consist of several
reactions chained together. For example (1) sketches
the pathway for the transcription of the lac operon. The
transcription of the lac operon will be the leading
example.
transcription lactose: nucleotides + DNA_lactose ↔ …
↔ mRNA_lactose + DNA_lactose (1)
Formulae like (1) do not express inhibitors,
activators, speed and equilibrium conditions. For
example lactose and CRPcAMP are the activation
proteins regulating the transcription of the lac operon.
Within the well-known Michaelis-Menten equations the
rate of a reaction can be derived on the basis of
concentrations of substances, binding constants,
stoichiometry values and equilibrium constants.
Michaelis-Menten provides formulae for the reactions
in continuous time. Equations such as Michaelis-
Menten equations can be extended with inhibitors and
activators. Using these formulae, a complete description
of the processes in the cell could be given if all the
BDI Model
desire
reason
intention to prepare
reason
beliefs environment
intention to perform
belief in opportunity
action
reactions and their parameters were known, which is
not the case. Example parameter values are given for
the transcription of the lac operon reaction, see (2).
Regulators: lactose 0.01, CRP_cAMP 0.01,
kcat = 0.01, keq = 100. (2)
Viewed from a more abstract perspective, what does
this reaction do over time? When enough of lactose,
CRPcAMP and nucleotides are present, the
mRNA_lactose will start to be produced, and after a
certain delay a significant amount of mRNA_lactose
will be present. The concentrations of lactose and
CRPcAMP need to be sufficiently high for a certain
period of time in order for the reaction to proceed, a
concentration of at least 0.1 mmolair (the threshold) of
both is sufficient in the example. The amount of
nucleotides needed for the reaction to proceed is at least
about 0.1 mmolair again. A ready supply of nucleotides
is always synthesised by the cell. In order for the
reaction to happen, the amount of mRNA must not be
so high as to impede the reaction, a concentration lower
than about 10 mmolair in this example. When the
reaction proceeds, the amount of nucleotides will
slowly decrease. The amount of mRNA will slowly
accumulate by this reaction. Other parts of the system
will supply new nucleotides and the mRNA will
degrade after some time.
The large amount of unknown parameters, and
computational complexity of integrating the resulting
differential equations make a model using only
chemical differential equations unwieldy. Therefore a
more abstract description is introduced. The process is
modelled in our temporal environment as follows.
Temporal relationships are defined between a number
of sources and an effect. Parameters are used to specify
the minimal duration of the sources, the delay before
the effect becomes apparent, and the duration of the
effect; for the delay a minimum and maximum value
can be set. As an illustration, the temporal relationship
between the substances in the transcription of the
lactose operon is determined. Since nucleotides are
always present, these do not need to be mentioned in
the temporal relationship, as it does not influence
behaviour. The temporal relationship to determine
when the mRNA_lactose is produced is denoted as:
DNA_lactose + lactose + CRP_cAMP
•→
→e,f,g,h mRNA_lactose
On the left-hand side the conditions that have to be
met are listed. The DNA_lactose, meaning the presence
of the lactose operon in the DNA. Also lactose, meaning
the presence of lactose and CRP_cAMP (i.e.,
concentration above a threshold value), meaning the
presence of CRP_cAMP to bind to the activation sites
of the operon are listed on the left side. On the right
hand side, the change that will happen later is listed,
mRNA_lactose meaning the presence of lactose mRNA
that is produced. The parameters e, f, g and h are
positive real numbers that set the minimum and
maximum delay (e and f), the condition duration (g)
and the result duration (h). Realistic parameters for the
values of e, f, g and h for the example are e = 60, f =
60, g = 1 and h = 40, as the process to create the mRNA
takes about 60 seconds, and the mRNA will stay in
existence for about 40 seconds on average. When the
condition holds for 1 second or more, the transcription
process starts.
Previously, a temporal model has been presented of
chemical processes using categories of substances and
temporal relationships between these. Here the
temporal relation •→ → that is called the “leads to”
relation is more precisely defined. When α •→ →e,f,g,h β
this means that:
if property α holds for a while (g), then some time
(between e and f) later property β will hold for a
while (h).
The definition of the relationships as given above,
can be applied to situations where the sources hold for
longer than the minimum g. The result for a longer
duration of α for α •→ → β is depicted in Figure 2. The
additional duration that the source holds, is also added
to the duration that the result will hold, provided that
the condition e + h ≥ f holds. This is because the
definition can be applied at each subinterval of α,
resulting in many overlapping intervals of β. The end
result is that the additional duration also extends the
duration that the resulting notion β holds. More
formally:
[α •→→e,f,g,h β & e+h ≥ f] ⇒ ∀a≥0: α •→→e,f,g+a,h+a β
additional duration
duration g
additional duration
notion α
notion β
duration h
actual delay
minimum delay
maximum delay
Figure 2: Temporal relationships for longer durations.
Using these temporal relationships, the bacterial
regulation can be modelled from the chemical
perspective. The temporal relationships capture the
timing of the underlying chemical reactions. The
durations and delay minimum and maximum can be
specified to fit the timing of the chemistry. The
(3) formally defined temporal operator “leads to” aids the
construction of simulation and derivation software to
support the inspection of modelling results.
4. Temporal Modelling of
Intentional Dynamics
Formalised models for intentional notions like those of
(Rao & Georgeff, 1991) do not take into account their
dynamics. To be able to closely relate an intentional
model to the bacterial embodiment in chemical
processes, such dynamics is crucial. Therefore a
temporal modelling approach to intentional dynamics is
introduced, based on the temporal “leads to” relation
introduced in the previous section. The dynamics of
intentional notions is expressed in terms of this “leads
to” relation. By applying the correspondences between
the intentional notions and the chemical substances
from the previous sections a dynamic model based on
intentions is obtained. The resulting model is a correct
and transparent high level description of the regulation
process, understandable for the reader not versed in the
technicalities of the chemical pathways in the cell.
The model presented here covers food import
behaviour of E. Coli. The temporal relationships
between desire(lactose_import) and intention(lactose_import),
using belief(lactose_externally_present) and belief(not
glucose_externally_present) are discussed here as an
example. The desire and the beliefs (the reason for the
creation of the intention) must hold for at least some
duration. After a delay larger than the minimum delay
and shorter than the maximum delay, the intention
starts to hold for some duration. This temporal
relationship is denoted in relation (4).
desire(lactose_import) ∧ belief(lactose_externally_present) ∧
belief(not glucose_externally_present)
•→→e,f,g,h intention(lactose_import). (4)
The intentional notions are related to the substances,
as discussed in the Sections 2.3 and 3.1.
In relation to (4), the lactose and CRPcAMP
substances are interpreted as beliefs. The DNA relates
to a desire and the mRNA to an intention. The
nucleotides and other, intermediate, substances are not
labelled with intentional notions. These substances are
only the machinery of the embodiment of the bacterial
cognition, and play no decisive role in the lactose
uptake behaviour. This means that, leaving out these,
the intentional model provides a more abstract picture
of the processes. If new insights were to prove some
substances play a significant role in the decision
process, these can easily be added. Assigning the
intentional notions to the substances is not enough. It is
also necessary to know at which concentration of the
substance the intentional notion holds. A threshold is
used to determine whether the intention notion holds or
not.
 The timing parameters e, f, g, and h are the same as
those found in the abstract chemical model, thus
relation (5) holds.
desire(lactose_import) ∧ belief(lactose_externally_present) ∧
belief(not glucose_externally_present)
•→→ 60,60,1,40 intention(lactose_import). (5)
Some more example intentional temporal
relationships within the model are:
................................ Desires ................................
desire(grow).
desire(food_import).
desire(lactose_import) .
desire(glucose_import).
The cell always desires to grow. From this basic desire
stem the other desires, which also always hold. The cell
desires to import nutrients (in order to grow). The cell
also desires to import glucose (in order to import
nutrients), and to import lactose (in order to import
nutrients).
.......... Intentions to prepare import actions ..........
desire(lactose_import) and additional_reason1.1 •→
→ a large variety of physical and chemical processes are
60,60,1,40intention(prepare_lactose_import) .
desire(glucose_import) and additional_reason1.2 •→
→ moderately complex organisms, abstraction and
60,60,1,40 intention(prepare_glucose_import) .
additional_reason1.1 = def belief(lactose_externally_present)
and belief(not glucose_externally_present).
additional_reason1.2 = def true.
The cell will intend to prepare to import a nutrient if
sufficient additional reasons are present. The desire to
import lactose, combined with the additional reason to
import lactose (additional_reason1.1) results in the
intention to prepare to import lactose. The additional
reasons to import lactose are the belief that lactose is
present outside and the belief that glucose is not present
outside. The desire to import glucose, when combined
with an additional reason (additional_reason1.2), results in
the intention to prepare to import glucose. There is no
additional reason needed to import glucose, it is
denoted true.
.......... Intentions to perform import actions ..........
intention(prepare_lactose_import) and additional_reason2.1
•→
→ 0,0,60,600 intention(perform_lactose_import).
intention(prepare_glucose_import) and additional_reason2.2
•→
→ 0,0,60,600 intention(perform_glucose_import).
additional_reason2.1 = def belief(lactose_externally_present).
additional_reason2.2 = def belief(glucose_externally_present).
Given some good additional reasons and the intention
to prepare an import, the cell will generate the intention
to perform the import. When the cell intends to prepare
lactose import, and enough additional reason is present
(additional_reason2.1), the belief that lactose is present
outside, then the intention to perform the import of
lactose is generated. When the cell intends to prepare
glucose import, and enough additional reason is present
(additional_reason2.2), the belief that glucose is present
outside the cell, then the intention to perform the
glucose import is generated.
5. Discussion
The relationship between the chemical regulation
substances and the intentional notions for the behaviour
description shows that the intentional model presented
in Section 4 is grounded by the chemical processes of
the regulation. The simulation of the intentional model
proves that the intentional model corresponds to the
chemical bacterial regulation. In other words, the BDI
model apparently matches well with the regulation that
happens in living cells.
The value of this work for Biology lies in managing
the complexity of living systems. For example, the
internal processes within organisms often are so
complex that explanations of their behaviour in terms of
inaccessible. This paper shows how, at least for
intentionalisation of such continuous processes can be
done in a justifiable manner. The resulting models show
intentional dynamics embodied in physical and
chemical models of real world dynamics.
References
Clark, A. (1997). Being There: Putting Brain, Body and
World Together Again. MIT Press.
Clark, A. (1999). Where brain, body, and world collide.
Journal of Cognitive Systems Research, 1, pp. 5-17.
Dennett, D.C. (1987). The Intentional Stance. MIT
Press. Cambridge Mass.
Dennett, D.C. (1991). Real Patterns. The Journal of
Philosophy, vol. 88, pp. 27-51.
Neidhardt, F.C., Curtiss III, R., Ingraham, J.L., Lin,
E.C.C., Brooks Low, K., Magasanik, B., Reznikoff,
W.S., Riley, M., Schaechter, M.& Umbarger, H.E.,
eds. (1996). Escherichia coli and Salmonella
typhimurium. ASM Press, Washington, D.C.
Rao, A.S. & Georgeff, M.P. (1991). Modelling Rational
Agents within a BDI-Architecture. In: J. Allen, R.
Fikes & E. Sandewall, eds., Proceedings of the
Second International Conference on Principles of
Knowledge Representation and Reasoning, (KR’91),
Morgan Kaufmann, pp. 473-484.
Sun, R. (2000). Symbol grounding: a new look at an old
idea. Philosophical Psychology, 13, pp. 149-172.
 



!"""
#%$&' ')(

"+*,
-. /
-&01
!2
3(4
568792:<;>=?A@,B
CEDGFHJIAK @L=3MN68BO6 D ?P;Q?JRPST;Q7UV68?A@WX? K @,Y8Y86[ZQ@,?V7\@]^;>_%`AaN@ HVKLbECEDGFHJIAK @L=cd7L68@,?V7\@;Q?VRTc DQefKhg ;>=O@ij?AZG68?A@L@L=68?AZA`
k UA@l?V6[mQ@L=BO6 Khn2DQe So@,Y[_ DGI =?A@Q`dM687 KpD =68;d`rq>sAtLsd`ruvlwc k3x u!]yWzu
? F _4{N7LB b|F}Ib~DQQb ; I
€2^‚Lƒ…„…†V‡Gƒ
a D 7O9d68?AZ68B; eˆI ?VRV; F @,? K ;QY^=O@,‰ I 6[=O@ F @,? K-efD =w; FD _J68Y[@= D _ DQKKpD _@};>_JY[@ KpD 68? K @L=;Q7 K!g 6 K U
D _dŠp@,7 K B68?‹6 K B@,?mŒ6[= D ? F @,? KLb Wz? K UV6B H ; H @L=Ž` g @ H =O@,Bp@,? KNK UA=O@L@"@ F _ D RV68@,RT; HVH = D ;Q7UA@,B KpDPK UV@
R D 7O9d68?AZ H = D _JY[@ FeD =2md68BO6 D ?d‘Z I 68RA@,R F"D _r68Y[@’= D _ DQK B b k UA@T; HJH = D ;Q7UV@,B“@,;Q7U”UV;,mQ@‹;•RV6—–@L=p
@,? KP˜ ;,m DGI =Ž`;Q?VR I BO@Q™•UV68ZGUdhY[@LmQ@,YmŒ6BO6 D ?šwY D Z> HD Y8;>= DQHVK 67L;QY ˜JDŽg š;Q?JR`-›AœA; K 6 D ?”;Q?VRŠ D 68? K
;Q?AZGY[@,B bŸž @’;QY8B D H =O@,Bp@,? K ; F @ K U D R D Y D Z nefD ="RA@LmQ@,Y DQH 68?VZo@ F _ D RV6[@,R•; HVH = D ;Q7UA@,B KpD¡F"D _r68Y[@
= D _ DQK ?V;,md6[ZG; K 6 D ? H = D _JY[@ F B,` D ? K UA@N_J;QB68B DQeyg UV67U;QY8Y K UA=O@L@ F @ K U D RVB g @L=O@wRA@LmQ@,Y DQH @,R b k UJ68B
F @ K U D R D Y D Z n‹F ;>9Q@,B_ DQK U‹@\œ H Y867L6 K ;Q?VRŸ6 F"H Y687L6 K!I Bp@ DQe @ F _ D RJ6 F @,? K 6?Ÿ7L; K @LZ D =68BO68?VZPmŒ68B I ;QY
RV; K ; KpD¢e ;Q7L68Y6 K ; K @@L£¢7L6[@,? K3H @L=7\@ HVK 6 D ?P;Q?VR’;Q7 K 6 D ? b
¤¥ (4
#¦(

§¨d©ªo¨d«¬­f®w©&¨Œ«^¨Œ¯&°²±r³¨r´LµV¬¶·2¬°N¸“¹‘º¶4³¸Œªo®²¶A¯L¸Œ­y©&®…»Vº¬f©®²ªo®²¶A¯¯&¨«^®¸Œ«­®¯&¨¼^®…©¹‘¨d©ª½¼©&®…´²¬f°®¬f¶A¯&®²©\¸´Ž¾
¯\¬¨d¶°²¿Àp¶³4º°¯&©&¬ˆ¸­E¹‘¨d©&µd¾X­f¬¹Á¯&°²±yªT¨Œ«¬f­®“ªT¸Œ¶Vº4¹h¸Œ´Ž¯\º©¬f¶·Ÿ¸°&°&®…ª«­Âé&¨Œ«^¨Œ¯&°²±)¸¶³Ä©&®…°&´²º4®o©¨d«y¨Œ¯\°"¸­f­
¶®²®…³
¯\¨ªT¨QŌ®’¯&¨¸Œ¶Æ¨d«JÇ®²´Ž¯2¨¹¬¶V¯&®²©®²°¯“¸Œ¶4³È³¨J´Lµ•É¬¯\ÊƬ¯"¯&¨´²¸Œ©&©ÂĨdº4¯"¯&Ê®²¬©¯L¸Œ°µV°²¿oÀp¶Ë¯&ʬf°"¼¸¼^®…©²±)É®
®ŽÌ4¸Œªo¬f¶®“¯\Ê©®²®’°&®…¼¸Œ©\¸Í¯\®P®²ª2«^¨J³¬f®…³Ãªo®…¯\Ê4¨r³°"³4®…Åd®…­f¨d¼y®²³Æ«AÂï&Ê®’¸Œº4¯&ʨd©"¸Œ¶4³Ä´²¨d­­f®>¸·dº®²°}¯\¨ŸÊ¸Œ¶4³­f®
¯\Ê® ³¨r´LµV¬¶·Ã¼©¨d«­f®…ªÎ¹‘¨d©2·d©¨dº¶³J¾p«¸Œ°®²³©¨d«y¨Œ¯\°…¿ÃÏ-¸Œ´Lʔ°ÂJ°¯&®²ªÐª ¸Œµ®²°®,Ìr¼­f¬´²¬¯“º°®‹¨Œ¹w©¨d«^¨¯®²ªP¾
«^¨J³¬ªT®…¶V¯¬¶•¨d©³®²©¯\¨T¹h¸´²¬f­¬¯\¸¯\®©&¸Œ¼¬³•®…Ño´…¬f®²¶A¯³¨r´LµV¬¶·¿§¨d©}³4¬Ò^®²©&®…¶A¯©®>¸Œ°¨d¶°²±^®>¸Œ´LÊĨ¹E¯\Ê4®“¯\Ê4©&®²®
¸Œ¼¼©¨A¸Œ´LÊ®…°³®²¼y®²¶³Æ¨d¶Ä¯&Ê®º°®o¨¹-¸Œ¶Æ®²ª«y¨r³4¬f®²³Ã¸Œ¼¼©¨A¸Œ´LÊÓ¿¢Ô­¯\ʨdº4·dÊï\ʬ°¼¸Œ¼y®²©}¹‘¨r´…º°&®…°"¨d¶Ä¯\Ê®
³¨J´LµJ¬¶·¼©¨d«­®²ª±r¬¯®,̸ªT¬¶®²°!¬f°&°º®²°!¯\ʸ¯¸Œ©®©&®…­f®ŽÅŒ¸¶V¯!¯\¨P¸­f­%©&®…°&®>¸©&´Lʟ¨d¶‹ªo¨d«¬­f®©&¨Œ«^¨Œ¯&°²¿<ÕÈ®¸Œ­°&¨
¼©&®…°&®…¶V¯¸P·d®²¶®…©\¸Œ­)ªo®…¯&ʨJ³¨d­f¨Œ·ŒÂT¹‘¨Œ©w³®…ÅŒ®²­f¨Œ¼¬f¶·’®…ª«^¨J³¬®²³¡ªT¨Œ«¬f­®©&¨Œ«^¨Œ¯&¬f´°ÂJ°¯\®…ªT°…¿
Ô­­w¨Œ¹¯&Ê®²°®Ö°Âr°¯\®…ªT°Tʸ>Åd®«^®…®²¶×³4®…Åd®…­f¨d¼y®²³×º¶³®…© ¸ÆªT®Ž¯\ʨJ³¨d­¨d·ŒÂȨŒ¹®…ª«y¨r³¬®²³Ø¼y®²©&´…®²¼4¯&¬f¨d¶Ù¹‘¨d©
ªT¨Œ«¬f­®P©&¨d«y¨Œ¯¢·dº¬³¸Œ¶´…®Œ±Éʬf´LÊËÉN¸Œ°¢Ú©&°¯2¼4©&®²°®²¶A¯\®…³”¬f¶×ÛÝÜGÞz±j¸Œ¶³È¬f°®ŽÌr¼¸Œ¶4³®²³¬¶Ë¯&ʬf°¢¼¸Œ¼y®²©…¿‹ÀX¯“¬f°
É®…­f­8¾pµV¶¨Qɶ߯\ʸ¯“¯&Ê® ³¸¯L¸¡¹‘©&¨dªà¸Ö°¬f¶·Œ­f®’¬fªT¸Œ·d®’¬f°¢¬f¶¸Œ³®²»Vº¸Í¯\®T¯&¨•¬¶V¯&®²©¼©&®Ž¯P¸¯&Ê©&®…®Ž¾X³¬fªo®²¶°¬f¨d¶¸Œ­
°&´…®²¶®¿”á}¨QÉ!®…Åd®…©²±-®…ÅŒ®²¶Ù¬¹¬¯¬f°2¼y¨d°°&¬f«4­f® ¯&¨Ä©&®²´…¨QÅd®²©P¯&Ê®‹°&´…®²¶®¡°¯\©&º4´…¯\º4©&®Œ±-º°¬f¶·Äªo®…¯&ʨr³4°°&º´LÊ+¸Œ°
°¯&®²©&®…¨‹¸Œ¶4³°¯\©º´…¯&º©&®¹‘©&¨dªâªT¨Œ¯&¬f¨d¶¦±4¸Œºr¯\ʨd©°°&º´Lʕ¸Œ°ãº¼©&®TÛåäÍÞʸ>Åd®¼y¨d¬f¶A¯&®²³¨dº4¯w¯\ʸ¯¯\Ê®…©&®¬f°w¶¨
º¶¬»Jº4®¢°ÂJ°¯\®…ªæ¨Œ¹E´²¸¯\®…·d¨d©¬f°\¸Í¯\¬f¨Œ¶‹¯\ʸ¯¸­f­f¨QÉ°w­f¸Œ«^®…­f­¬f¶·“¨Œ¹j¯&Ê®®²­®²ªo®²¶A¯\°¨Œ¹
¯\Ê®°´²®²¶4®Œ¿
Ï3ª«^¨J³¬ªT®…¶V¯¬°w¯\Ê®¯\Ê®…¨d©Ÿ¯&ʸ¯}¸Œ«°¯\©&¸Œ´…¯ªo®>¸Œ¶¬¶·±©®>¸Œ°¨d¶•¸¶³•¬ª ¸Œ·Œ¬f¶¸¯&¬f¨d¶Tʸ>Åd®2¸o«y¨r³4¬f­ÂT«¸Œ°¬f°
Û8ç>èÞz¿“Ï3ª«^¨J³¬ªT®…¶V¯¬f°¹‘¨Œ©&ªo®²³Ä«AÂï\Ê4®P¶¸¯\º4©&®P¨Œ¹-ÊVºªT¸Œ¶Æ«^¨J³¬f®…°²±Ó¬¶´²­º³¬¶·‹¼y®²©&´…®²¼4¯&¬f¨d¶Ë¸Œ¶4³ÄªT¨Œ¯&¨d©
ªT¨QŌ®²ªo®²¶A¯>±y¸Œ¶³¨Œ¹E¨dº©¼ÊAÂJ°&¬f´²¸Œ­
¸Œ¶³°&¨J´²¬f¸Œ­)¬f¶A¯\®…©\¸Œ´Ž¯\¬¨d¶°Nɬ¯&ʝ¯&Ê®"ɨŒ©&­f³¦¿!Ô}º4¯\ʨŒ©&°°&º´LÊĸ°}鸌µ¨ŒÒ
Û8çŒç²Þ
¸Œ©&·Œº®¯\ʸͯ}ÊVºªT¸Œ¶¡´>¸Í¯\®²·Œ¨d©&¬°\¸¯&¬f¨d¶Ÿ°Âr°¯\®…ªT°}¸Œ©&®"®…ª«y¨r³¬®²³Ó¿
Àp¶Ã¸’°&¬ªT¬­ˆ¸Œ©!ª ¸Œ¶4¶®²©wÉ!®2°&º4·d·d®²°¯}¯&ʸ¯©&¨Œ«^¨Œ¯&¬f´´>¸Í¯\®²·Œ¨d©&¬°\¸¯&¬f¨d¶Æêh«^¨Œ¯&ÊÖ®,Ìr¼­f¬´²¬¯¸¶³•¬ªT¼­¬f´…¬¯…ëE°&Ê4¨dº­f³
«^®•®²ª«y¨r³4¬f®²³¦¿ìÀp¶ÛåÜGÞí±É®Ä¼©¨d¼y¨d°&®…³î¯\ʸ¯Ÿ¸ªT¨d«4¬f­f®¡©¨d«^¨¯>ï尋®²ª«y¨J³¬fªo®²¶A¯‹´²¸Œ¶ð«y®Ãº°&®Ž¹‘º­f­Â×°®²®…¶
¬f¶¯&Ê©&®…®¡¸Œ°&¼y®²´Ž¯\°²ñ ¸Ã¼ÊAÂJ°&¬f´²¸Œ­N©¨d«^¨¯\¬f´T«^¨J³4ÂyòN¸•¯\¸Œ°&µyêh°Žë¢¬f¶ÈÉʬf´Lʔ¯\Ê4®‹©&¨d«y¨Œ¯P¬°¬f¶AŌ¨d­ÅŒ®²³Óò¸Œ¶³Ó±3¸Œ¶
®²¶AÅJ¬f©¨d¶ªo®²¶A¯Qêh°Žë¦¬f¶Éʬf´LÊ¢¯\Ê®-©&¨d«y¨Œ¯¬°
¼y®²©¹‘¨d©&ªo¬f¶4·¯&Ê®3¯L¸Œ°µy¿Eó!´²¨Œ¶°&¬³®²©¬f¶·w¯\Ê4®²°&®¸Œ°&¼y®²´Ž¯\°…±d´²©®>¸¯&¨d©&°
¨Œ¹4ÅJ¬°&¬f¨Œ¶r¾p·dº4¬f³®…³"©&¨d«y¨Œ¯°ÂJ°¯\®²ªo°
´>¸Œ¶¹‘¨d©&ªô´>¸¯&®²·d¨Œ©&¬f°&¸¯\¬¨d¶"°ÂJ°¯\®²ªo°¯\ʸ¯¹h¸Œ´…¬f­¬¯L¸Í¯\®
©&¨d«y¨Œ¯
¸Œ´Ž¯\¬f¨Œ¶¢«¸Œ°&®…³
¨d¶³¸¯L¸¯&ʸ¯wÉ!¨dº­f³¨Œ¯&Ê®²©É¬f°&®"¹‘º¶4³¸Œªo®²¶A¯L¸Œ­­ÂTº¶³®…©&³®Ž¯\®…©&ªo¬f¶®"¯\Ê4®¢©¨d«y¨Œ¯>ïå°®²¶AÅJ¬f©&¨Œ¶ªT®…¶A¯>¿
õ%öhõÃ÷“ø
ùÄú
ûÓüGý)þ‘
ÿ 
û ‘
ù
ã¨r´LµV¬¶·’´>¸Œ¶¡«y®¢³4®…Ú¶®…³Ã¸Œ°wªo¨QÅJ¬f¶·¹‘©¨dªâ¯\Ê4®"´²º©©&®²¶A¯}¼^¨Œ°&¬¯&¬f¨d¶T¯\¨o¸P³®²°¬f©&®…³Ö¼y¨d°¬¯\¬¨d¶¡¸Œ¶³¡¨d©¬f®…¶V¯\¸Í¾
¯\¬¨d¶Ó±4Éʬ­f®¹‘¨Œ­f­f¨Qɬ¶·¸P°\¸Í¹‘®¯\©\¸QÇ®²´…¯&¨d©ÂÆÛ8ç²Þí¿ wÊ®"Ú¶¸­)¼^¨d°¬¯&¬f¨d¶Ÿ¸¶³¡¨d©&¬®²¶A¯L¸¯&¬f¨d¶ ¨Œ¹j¯&Ê®©&¨d«y¨Œ¯wª2º°¯
ç
 «^®¸Œ³®²»Vº¸Í¯\®2¹‘¨d©¯\Ê®¢¯\¨d­®²©&¸Œ¶´²®…°©&®…»Vº¬f©®²³Ã«A•¯&Ê®2¼¸Œ©¯&¬f´²º4­ˆ¸Œ©¯L¸Œ°µ%¿
wÊ®°Âr°¯\®…ªT°¼4©&®²°®²¶A¯\®…³È¬¶Ö¯&ʬf°
¼¸Œ¼y®²©v¸©&®N©®²°¯&©&¬´…¯\®…³¯\¨·Œ©&¨dº¶4³r¾p«¸°&®²³Pªo¨Œ¯&¬f¨d¶Ó±GÅJ¬f®ŽÉÙ¯&Ê®N®…¶VÅJ¬©&¨d¶ªo®²¶A¯<¯&Ê©&¨Œº·dÊP´>¸ªT®…©\¸Œ°ê N¸Œ©¯\®²°¬ˆ¸Œ¶
¸Œ¶³Öé)¨d·¾ j¨d­ˆ¸©ŽëL±4¸Œ¶³³¨J´Lµ É¬¯&ʝ°¯L¸¯&¬f¨d¶¸Œ© ¨d«JÇ®²´Ž¯\°…¿
 /Ó(

"
!0
-  ì*,
!Öj 
 }
!!"" /
-&0Ë
-
3(Ö
 )} E(
!
wʬf°ªo®…¯&ʨr³4¨d­f¨d·Â}°º·d·d®…°¯&°E¸NɸQ¯\¨¸Œ¶¸Œ­ÂJ°&®3¸Œ¶³´…¨d¶°¯&©&º´Ž¯
®²ª«y¨r³4¬f®²³"ÅJ¬f°º¸Œ­d·Œº¬f³¸¶´²®v°Âr°¯\®…ªT°¹‘¨d©
ªT¨Œ«¬f­®<©&¨d«y¨Œ¯&°
¯&ʸ¯j¸Œ©&®3©¨d«º°¯>±Œ®ŽÑo´²¬f®…¶A¯<¸Œ¶4³“¯L¸°&µ"³¬©&®²´Ž¯\®…³Ó±·Œ¬Åd®…¶¸°&¼y®²´…¬Ú´-¼ÊAÂJ°&¬f´²¸Œ­V©&¨d«y¨Œ¯²±Í¯\¸Œ°&µyêí°,ëL±
¸Œ¶³Ö®²¶AÅJ¬f©&¨Œ¶ªT®…¶A¯Qêí°,ëL¿ wʬf°¬f°w¶¨¯¬¶V¯&®²¶³4®²³Ö¯&¨T«y®¢¸o³4®…Ú¶¬¯\¬Åd®"ÉN¸>¨Œ¹E¼©&¨J´²®…®²³¬¶·ò´²®…©¯\¸Œ¬f¶­ÂŸ¨Œ¯&Ê®²©
ªT®Ž¯\ʨJ³°-¸Œ©®¼y¨d°°&¬f«4­f®Œ¿Eá}¨QÉ!®…Åd®…©²±J¯\Ê®ªo®…¯\Ê4¨r³¨Œ­f¨d·ŒÂ“®²­º´²¬³¸¯&®²°3¸·d®²¶®…©\¸Œ­%¸¼¼©&¨d¸Œ´LÊ ¸¶³T³4®²°&´…©&¬«^®…°!¸
¼^¨Œ°&°&¬«­f®}¼©&¨J´²®…°&°…¿ wÊ4®Ú©&°¯w¼¸Œ©¯N¨Œ¹)¯&ʬf°!ªT®Ž¯\ʨJ³¨d­¨d·ŒÂ’³¬©&®²´Ž¯\°w¸¶¸Œ­ÂJ°¬f°!¯&¨2¯\Ê4®"¸Œ¼¼©¨d¼©¬ˆ¸¯&®¸Œ°¼^®…´…¯\°
¨Œ¹j®²ª«y¨r³4¬fªo®²¶A¯¬f¶Ÿ´…¨d¶°¯&©&º´Ž¯\¬¶· ¸Pº°®…¹‘º­´²¸¯\®…·d¨d©&¬°\¸¯&¬f¨d¶‹¹‘¨Œ©N¯\Ê®°ÂJ°¯\®²ª¿
çd¿ wÊ®  ý ³®ŽÚ¶®²°Éʸ¯´²¸¯\®…·d¨d©&¬®²°¸Œ©®©&®²»Vº¬©&®…³Ó¿
¿ wÊ® j"ø !  þhü  
ù #3ûÓú
$þ ßùVÿ T¨Œ¹3¯\Ê4®©&¨d«y¨Œ¯²±)¸Œ­f¨Œ¶·Tɬ¯\Êï\Ê4%®  ý ³®ŽÚ¶®…°¯&Ê®ÅJ¬f®ŽÉ°"¨Œ¹3¯\Ê®

®²¶AÅJ¬f©¨d¶ªo®²¶A¯}¸Œ¶³¨d«JÇ®…´…¯\°¬f¶Ÿ¬¯w¯\ʸͯw¯\Ê®©&¨Œ«^¨Œ¯ªT¸>Ÿ®²¶4´²¨dº¶A¯&®²©²¿
&  ¿ 
¨2Ú¶³ ÉÊ4¬f´LÊ ¹‘®²¸¯\º©®²°w¸©&®}¸Œ³®²»Vº¸Í¯\®¯&¨2³¬°&´²©¬fªo¬f¶¸Í¯\®w´>¸Í¯\®²·Œ¨d©&¬®²°²±J´²¨Œ¶°&¬³®²©-¯\Ê®…¬f©!¸Œ¼¼y®>¸Œ©&¸Œ¶´²®
¹‘©&¨Œªâ¯\Ê®"ÅJ¬f®ŽÉ°®²¶´…¨dº¶A¯\®…©&®²³•¬f¶‹¯\Ê4®¢©¨d«y¨Œ¯>ïå° ùJ(ÿ '))þ û%
ÿ ßùVÿ A¿
*}¨Œ¯&®¯\ʸͯ¨Œ¯\Ê®…©w¬f¶A¯\®…©\¸Œ´Ž¯\¬¨d¶°w¸Œ©®"¼^¨Œ°&°&¬«­f®±r¹‘¨d©®ŽÌ4¸Œªo¼­f®¸­f¸Œ´Lµ ¨Œ¹j¸>ŸŒ¬f­f¸Œ«¬­f¬¯p¨Œ¹
°&º¬¯L¸Œ«­®³¬°&´²©¬fªP¾
¬f¶¸Í¯\¬f¶4·“¹‘®>¸Í¯\º©®²°wªT¸QÂT­f®²¸Œ³‹¯\¨’¸“³¬Ò^®²©®²¶A¯´²­f¸Œ°&°¬Ú´²¸¯\¬¨d¶ °Âr°¯\®…ªÖ±¨d©®…Åd®…¶¡ªT¨J³¬¹ÁÂJ¬¶·“¯&Ê®¯L¸°&µ «¸Œ°&®…³
¨d¶Éʸ¯¼^®…©&´²®…¼4¯\º¸Œ­N¬f¶r¹‘¨d©&ªT¸¯\¬¨d¶Ë¬f°P¸>ŸŒ¬f­f¸Œ«­®o©&¸¯\Ê4®²©2¯&ʸŒ¶Éʸ¯¬f°³®…°&¬©\¸Œ«­®Œ¿ÄÔ}­f°&¨4±<´LʸŒ¶·Œ®²°¯\¨
¯\Ê®“¼ÊAÂJ°&¬f´²¸Œ­
®²ª«y¨r³4¬fªo®²¶A¯ª ¸>«^®¢´²¨Œ¶°&¬³®²©®²³Ó±^°&º´LÊƸŒ°É©&¬¯\¬¶·T¶4®…É°&¨¹Á¯pÉN¸Œ©®ªo¨r³º4­f®²°w¯&¨ Ê¸Œ¶4³­f®
³¬Ò^®²©®²¶A¯¯pÂJ¼y®²°N¨¹)³¸¯\¸4±r¨Œ©N¸Œ³³4¬f¶·“¶®…É.°®²¶°¨d©&°N¯&ʸ¯!«^®Ž¯&¯\®…©®,Ì4¼4­f¨d¬¯³¬Ò¦®…©&®…¶V¯&¬ˆ¸¯&¬f¶·"¹‘®²¸¯\º©®²°…¿3Õ.¬¯\Ê
¯\Ê®…°&®¢¼^¨Œ¬f¶A¯\°w¶¨¯\®²³¦±¨dº©°&º·Œ·d®²°¯\®²³ÖªT®Ž¯\ʨJ³¨d­¨d·ŒÂT¬f°¸Œ°w¹‘¨d­­f¨QÉ°²ñ
çd¿wÔ¶¸Œ­ÂJ°&®"¯&Ê+®  ý , .-}¯&¨’¬f³®…¶V¯&¬¹ÁÂT¯\Ê®©&®…»Vº¬f©®²³¼^®…©&´²®…¼4¯\º¸Œ­)´>¸¯&®²·d¨Œ©&¬f®…°²¿
¿wÔ¶¸Œ­ÂJ°&®N¯&Ê®N¼y¨d°&°¬f«­®N¬f¶A¯&®²©\¸´…¯\¬¨d¶°
¯\Ê®w©¨d«^¨¯<ªT¸Q“ʸ>Åd®Nɬ¯\ʯ&Ê®w®²¶AÅJ¬f©¨d¶ªo®²¶A¯3¸Œ¶³P¨Œ«JÇ®²´Ž¯\°3¬f¶
¬¯²±%®ŽÌ4¸Œªo¬f¶¬¶·T«y¨Œ¯&Ê j"ø !  þhü  /ù #3û)ú0þ ÈùVÿ o¸Œ¶1³ 2 ý , 3-¯\¨T¬f³®…¶A¯\¬¹Á ¯\Ê®¢År¬®…É°¯\Ê®“©&¨d«y¨Œ¯
ªT¸Q‹¼^®…©&´²®…¬ÅŒ®¹‘©¨dªâÉʬ´LÊÖ¬¯wªº°¯³¬f°´²©&¬ªT¬¶¸¯&®}¯\Ê4®¢©®²»Vº¬©&®²³¼y®²©&´…®²¼4¯&º¸Œ­´>¸Í¯\®²·Œ¨d©&¬®²°²¿
& ¿wÔ¶¸Œ­ÂJ°&®“¯\Ê®P¬¶°¯\¸Œ¶´²®…°¨Œ¹3¯\Ê4®´>¸¯&®²·d¨Œ©&¬f®…°¨Œ¹-°¯L¸Œ·d® çd±^¹‘©¨dª¯&Ê®2ÅJ¬f®ŽÉ°¹‘¨dº¶³•¬f¶Ã°¯L¸Œ·d® ±Ó¸Œ¶³
Ú¶³•¸P°&®…¯}¨Œ¹
¹‘®>¸¯&º©&®…°¯&ʸ¯¬°®ŽÒ¦®…´…¯&¬Åd®“¸Œ¶³®…Ño´…¬f®²¶A¯¬f¶¡¼y®²©´²®²¬ÅJ¬f¶·’¯\Ê®©®²»Vº¬f©®²³¬f¶4¹‘¨Œ©&ªT¸¯\¬¨d¶
¬f¶Ÿ¯&Ê® ùJ(ÿ '))þ û%
ÿ ßùVÿ  , .-}ÉÊ4®²©&®¯\Ê4®¯\¸Œ°&µyêh°Žë!¬f°N¯&¨T¨J´²´…º©²¿
4 ¿wÔ¶¸Œ­ÂJ°&®T¸Œ¶AÂĬ¶V¯&®²©&¸Œ´…¯&¬f¨d¶°"¯&ʸ¯¢®²ªo®²©·d®P¯\¨¡­f¨J¨dµÖ¹‘¨Œ©¢´LʸŒ¶·d®…°“¨Œ¹´>¸¯&®²·d¨Œ©&¬f®…°¯\ʸͯ“ªT¸>ÂÄ«y®o©®Ž¾
»Vº¬f©®²³Ö¸Œ¶³°¯\¸Œ©¯¸Œ·d¸Œ¬f¶¡¸¯°¯L¸Œ·d®P碬¹
¶®ŽÉì´>¸Í¯\®²·Œ¨d©&¬®²°¸Œ©®¹‘¨Œº¶³Ó¿
wÊ®"©&®…ª ¸¬f¶³®…©!¨¹)¯\Ê®¼¸Œ¼^®…©N´²¨Œ¶°&¬³®²©°¯\Ê©®²®¸Œ¼4¼©&¨A¸´LÊ®²°¯&¨¯\Ê®}³¨J´LµJ¬¶·P¼©¨d«­f®…ª ¯\ʸ¯Ê¸>Åd®"«y®²®…¶
´²¨d¶4°¯\©º´…¯&®²³Ãº4°&¬f¶4·’¯&ʬf°wªo®…¯&ʨJ³¨d­f¨Œ·ŒÂd¿
56 /
"
0p+  vr
"!
Œ0)!
0’}4
! "
#%$&½7 J(4

Àp¶+9Û 8Þí±)¸Œ¶Ä®…ª«y¨r³¬®²³Æ¸Œ¼4¼©&¨A¸´LÊÃÉN¸Œ°"º4°&®²³Æ¯\¨Ÿ©®²³®ŽÚ¶®¯&©\¸Œ³¬¯\¬¨d¶¸Œ­)ÅJ¬f®ŽÉ®…©¾p´…®²¶A¯\©®²³È©&®²¼4©&®²°®²¶A¯L¸Í¯\¬f¨Œ¶°²±
ÂJ¬f®²­³¬f¶4·Ä´²¨dªo¼º4¯\¸¯\¬¨d¶¸Œ­w¸¶³+©®²´…¨d·d¶¬¯\¬f¨Œ¶Ø¸Œ³4ŸŒ¶A¯L¸Œ·Œ®²°²: ¿ wÊ®¡©&¨Œ«^¨Œ¯’°ÂJ°¯\®…ª ÉN¸Œ°T¸Œ«­®‹¯\¨Ë¬f³®…¶A¯\¬¹ÁŒ±
¶¸>ÅJ¬f·A¸¯&®‹¸Œ©¨dº¶³¦±v¸Œ¶³³4¨r´LµÆɬ¯&ʵV¶¨Qɶ”¨d«JÇ®²´Ž¯\°¡êh°&®²®‹§E¬f·dº4©&®çëL;¿ wÊ®‹º°®‹¨Œ¹¸Œ¶ß®²ª«y¨r³4¬f®²³¸Œ¼r¾
¼©&¨d¸Œ´LÊ¡¸Œ­f­¨QÉ®…³‹ªT¨J³®…­¾p«¸Œ°&®…³ ¨d«JÇ®…´…¯w©®²¼©®²°®²¶A¯L¸¯&¬f¨d¶4°¯\¨«y®°¬fªo¼­f¬Ú®²³T¸Œ¶³‹¨Œ¼4¯\¬ªT¬°&®²³o¹‘¨d©!¯&Ê®}¯\¸Œ°&µ%±
®²¶AÅJ¬f©¨d¶ªo®²¶A¯>±%¸Œ¶4³Ö¼ÊAÂJ°&¬´>¸Œ­®…ª«^¨J³¬ªT®…¶V¯N¨Œ¹j¯&Ê®©&¨d«y¨Œ¯²±¸Œ¶³¡Ê®…¶´²®¢ªT¨Œ©&®"®…Ño´…¬f®²¶A¯¸Œ¶³¡¼©&¸Œ´…¯&¬f´>¸­z¿
 <>=@? <>A.? <CBD? <CE3? <GFH?

§E¬f·Œº©&®oçdñÀpªT¸Œ·d®…°¯\¸ŒµŒ®…¶Æ¸Œ°¯\Ê®“©&¨d«y¨Œ¯}¶¸>År¬·A¸¯&®²°}¸Œ©&¨dº4¶³Ã¸o¼^¨QÉ!®²©°º¼¼­ÂdI¿ wÊ®2°ÂJ°¯\®…ª ©®²´…¨d·d¶¬°&®²°

¯\Ê®¨d«JÇ®…´…¯²±NªT¨QŌ®²°T¸Œ©&¨dº4¶³+¬¯Pɬ¯&Ê+µV¶¨QÉ­f®…³·d®¨Œ¹}¬¯\°’©®²­ˆ¸Í¯\¬ÅŒ®Ÿ¼^¨Œ°&¬¯&¬f¨d¶¦±!¸¶³×´²¸Œ¶+³4¨r´LµÉ¬¯\Êد\Ê®
©&®…»Jº4¬f©&®…³Ö°º©¹h¸´²®Œ¿
J 2 %ù õLK wÊ®…©&® ¸Œ©&®P¯\Ê4©&®²®o´>¸Í¯\®²·Œ¨d©&¬®²°"¯\ʸͯ2¸©&®’¬fªo¼^¨Œ©¯L¸¶V¯¹‘¨d©"¯\ʬ°¯\¸Œ°&µ%ñ¯\Ê®o¨d«JÇ®…´…¯“¨Œ¹!¬f¶A¯\®…©&®²°¯>±
¨d«°¯L¸Œ´…­f®²°…±¸Œ¶4³ ¹‘©&®…®Ž¾p°¼¸Œ´…®Œ¿ J 2 %ù1MNK wÊ®©&¨Œ«^¨Œ¯¬f°-·d©&¨Œº¶³r¾X«¸Œ°&®…³Ó±¸¶³‹Ê¸Œ°N¸“Ú4Ìr®²³ŸÊ®…¬f·dÊA¯´>¸Œªo®²©&¸4±
¯\ÊVº°…±E¬¹!¯\Ê®’¨d«VÇ®²´…¯2¬f°¢¸Œ°°&ºªo®²³Ë¯\¨«^®’°&¬¯&¯&¬f¶·¡¨d¶Æ¯\Ê®o·d©¨dº¶³Ó±j¸Œ­­<År¬®…É°¢¨Œ¹¯&Ê®o¨d«JÇ®²´Ž¯¸©&®Pɬ¯&ʬf¶
¸¼4­ˆ¸Œ¶®P¿ O«JÇ®²´Ž¯\°¢¨Œ¹N¸Ÿ°¬fªo¬f­ˆ¸©°&C¬ Q²®P¯\¨Ÿ¯&Ê®’©&¨d«y¨Œ¯¢¸Œ©®Pº¶­f¬µŒ®…­Â•¯&¨Ö«y®ÅJ¬f®ŽÉ®…³È¹‘©&¨Œª º¶³4®²©&¶4®>¸¯&Êߨd©
¸Œ«y¨GŌ®Œ¿Ö§¨d©“ªo¨d°¯2°º©¹h¸Œ´…®²°¨Œ¶­Â˸´²¨A¸Œ©°&®TªT¨J³®²­3¬°“©®²»Vº¬f©®²³Ó±E«º4¯2¸•ªT¨Œ©&®’³®…¯\¸Œ¬f­®²³ÈªT¨J³®²­3¨Œ¹¯\Ê®
³¨J´LµJ¬¶·P°&º4©¹h¸Œ´…®"ª ¸>Âo«^®¶®²´…®²°&°&¸Œ©Ÿ¸Œ°¯\Ê4®©&¨d«y¨Œ¯ªT¨QÅd®…°w´²­¨d°&®¿ wÊ®"­f®ŽÅd®²­^¨Œ¹³®Ž¯L¸Œ¬­¦©&®…»Vº¬f©®²³Ÿ´>¸¶Ÿ«^®
»Vº¸Œ¶A¯\¬Ú®²³‹«AÂP¯&Ê®}©¨d«y¨Œ¯-¬fªT¸Œ·d®²°<¯\¸ŒµŒ®…¶ ¹‘©&¨Œª¯\Ê®©&®…»Vº¬f©®²³ ¼¸¯\ÊÓ¿<§%º4©¯\Ê4®²©²±VÚ¶¬¯\®´>¸ªT®…©\¸©&®²°¨d­fºr¯\¬f¨Œ¶Ó±
´²¨dª2«¬f¶®…³ßɬ¯&Ê߯\Ê4®T¹h¸Œ´Ž¯2¯\ʸ¯2¯&Ê®‹©&¨Œ«^¨Œ¯“«y¨r³4Â˪T¸>ÂȨQŌ®²©&ʸŒ¶·Ã¯\Ê4®‹´>¸Œªo®²©&¸•­f®…¶°“ɬ­f­3¼©®…Åd®…¶A¯P¯\Ê®
©&¨d«y¨Œ¯w¹‘©¨dªâÅJ¬f®ŽÉ¬f¶·P¯\Ê4®¢¨Œ«JÇ®²´Ž¯¹‘©¨dª/Ō®²©¡´²­f¨Œ°&®¼©&¨>Ìr¬fªo¬¯pŒ±¸Œ¶4³Ö­®>¸Œ³°w¯&¨T¸Ú¶4¬¯\®°´>¸Œ­®"¨Œ¹j¹‘®²¸¯\º4©&®²°
Éʬf´LÊ¡´²¸Œ¶‹«y®¨d«°®²©Ō®²³Ó¿ J  %ù#RNK §%¨Œ©!¯\Ê4®¼^¨QÉ!®²©w°º¼¼­Â ¨d«VÇ®²´…¯°Ê¨Qɶ¡¬f¶‹§j¬f·dº©®2çd±r®²³·Œ®ªT¨J³®…­f°
É®…©&®"¸³®²»Vº¸¯&®}¯&¨2³¬°&´²©¬fªo¬f¶¸Í¯\®N¯\Ê4®¨d«JÇ®…´…¯²¿ j¨“©&®²´…¨d·d¶¬°&®}¨d«°¯\¸Œ´²­®²°…±4¸¢°&¬fªo¼­®ªo®…¯&ʨJ³T¨Œ¹Ó³®…¯&®²´…¯&¬f¶·
®²³·Œ®²°¨d¶Ã¯\Ê4%® S¨r¨Œ©ÉN¸°"º°&®…³Ó±)ɬ¯\ÊƸ¶V•°¯&©&¨d¶·¡®…³·d®²°¸°&°&º4ªT®…³Ë¯&¨Ÿ«^®P¸¶Ë¨Œ«°¯\¸Œ´²­®Œ±Ó°¬fªo¬f­ˆ¸©¯&¨Ÿ¯\Ê®
ªT®Ž¯\ʨJ³Ä¨Œ¹UÛ TÞz¿“§©&®…®Ž¾p°¼¸Œ´…®ÉN¸Œ°¢¸Œ¶A; S¨r¨Œ©}¯\ʸ¯"ÉN¸°"¶¨Œ¯"¨d«4°¯L¸´²­f®P¨Œ©¨d«JÇ®…´…¯>¿ J 2 %ùVWK wÊ4®P®²³·d®
ªT¨J³®…­f°"º°®²³ÆÉ®…©&®o»Jº4¬¯\®’°¬fªo¼­f®2¹‘¨Œ©"¬f¶³¬ÅJ¬f³º¸­E°&º4©¹h¸Œ´…®²°²±j¸Œ¶³ÄÉ!®²©®T¶¨¯2¸³®²»Vº¸¯&®P¯\¨¡·dº¸Œ©&¸Œ¶A¯\®…®T¸
º¶¬»Jº4®ªT¸¯\´LÊÓ¿EÀp¶‹¸“´…­fº4¯¯\®…©&®²³T®²¶AÅJ¬f©¨d¶ªo®²¶A¯>±4¸¢º¶4¬f»Vº®ªT¸¯&´LÊT¯\¨“¸“°º©¹h¸´²®¬°!³4¬Ño´²º­¯3¯\¨2·Œº¸Œ©\¸¶V¯&®²®
êí´…¨d¶°&¬³®²©T¯\ʮüy¨d°°&¬f«4¬f­f¬¯p”¨Œ¹¢¸ßªo¬f©©&¨d©o¬f¶Ù¯\Ê®•«¸Œ´LµV·d©&¨Œº¶³^ë\¿á}¨QÉ®ŽÅd®…©²±¸° ¯\Ê®•©&¨d«y¨Œ¯T¬f° ªo¨QÅJ¬f¶·
¸Œ©&¨Œº¶³Ö¯&Ê®¢¨d«JÇ®²´Ž¯>±¯\Ê4®“°Âr°¯\®…ª ®,Ì4¼4­f¨d¬¯\®²³¡°®…Åd®…©\¸Œ­
ÅJ¬f®…É°¨Œ¹E³¬Ò^®²©®²¶A¯°º©¹h¸´²®²°…±y¹‘º4°&¬f¶4·o¯&Ê®ª ¸¯&´LÊ®²°
¯\¨d·Œ®…¯\Ê4®²©wɬ¯&ʝ¨J³¨dªo®…¯&©&¬´¬f¶4¹‘¨Œ©&ªT¸¯\¬¨d¶Ó±VÉʬf´Lʝª ¸µŒ®²°ªo¬f°ª ¸¯&´LÊ®²°w­®²°°­¬fµ®²­ÂŒ¿
XZY â0 !<. [ !D0 vî [(\# 3^0 ]
_ 3a  ` bcE( ! *, !ð" #%$&
Àp¶Û & ± 4 Þz±NÉ!®Ã¼©®²°®²¶A¯\®…³.¸ß°ÂJ°¯&®²ª ¯\ʸͯ‹¬f°T¸Œ«­®¯&¨ß´…¨d¶A¯\©&¨Œ­}©¨d«^¨¯TÊ4®>¸Œ³¬¶·ß³4¬f©&®…´…¯&¬f¨d¶Ù¯\¨ß³4¨r´LµÙ¸¯
¸ËÚ4Ì4¸¯&®²³+¼^¨d¬¶A¯’«¸Œ°&®…³î¨d¶Ù¬fªT¸Œ·d®Ÿªo¨Œ¯\¬¨d¶Ó¿Ù§E¬8̸ͯ\¬f¨Œ¶ÙªT¸>”«y®¬¶³®…¼^®…¶³®²¶A¯ ¨Œ¹Ê®²¸Œ³¬¶·È³4¬f©&®…´…¯&¬f¨d¶
´²¨d¶A¯&©&¨d­í±¸Œ¶³Ç¨d¬f¶A¯¸Œ¶4·d­f®¬¶4¹‘¨d©ª ¸¯&¬f¨d¶ ¬f°w¶¨Œ¯©®²»Vº¬©&®²³¦±¨d¶­Â ­f¨d·Í¾p¼y¨d­ˆ¸Œ©!¨d¼4¯&¬f´>¸(­ S¨QÉ쬏°w©&®²»Vº¬©&®…³Ó¿
J 2 %ùÃLõ K §¨d©<¯&ʬf°<¯\¸Œ°&µ%±A¯\Ê®¨d¶4­ÂP´²¸¯\®…·d¨d©&¬®²°v¶®…´²®…°&°\¸©ÂT¸Œ©&®wÉÊ®Ž¯\Ê®…©-¯&Ê®´²º©©&®…¶V¯!Ê®²¸Œ³¬f¶4·¢³4¬f©&®…´…¯&¬f¨d¶
¬f°!­f®…¹Á¯¾p¨Œ¹¦¨d©-©&¬f·ŒÊV¯¾p¨Œ¹¦¯\Ê®Ú4Ì4¸¯\®…³‹¯L¸Œ©·d®…¯²d¿ wʬf°!¬f¶4¹‘¨Œ©&ªT¸¯\¬¨d¶’¬f°-º°&®…³Ÿ³¬f©®²´…¯&­ÂP¯&¨´²¨d¶A¯&©&¨d­^¸Œ³ÍǺ°¯\ªo®²¶A¯\°
¯\¨’Ê®²¸Œ³¬f¶4·o³4¬f©&®…´…¯&¬f¨d¶Ÿ¬¶Ö¸P´…¨dº¼­®²³¼^®…©&´²®…¼4¯\¬¨dL¶ eŒ¸´…¯\¬¨d¶Ÿ­f¨J¨d¼Ó¿ J  yù#MNK wÊ®·d©&¨dº4¶³r¾p«¸Œ°&®…³Öªo¨Œ¯\¬¨d¶
¨Œ¹3¯\Ê®P©¨d«y¨Œ¯Pêh¼ÊAÂr°¬f´²¸Œ­h븶³ÄÚ4Ì4¸¯\¬¨d¶Ã¨d¶Ã¯&Ê®P¨d«JÇ®…´…¯oêh¯\¸Œ°&µë¼­f¸Œ´²®2´²¨d¶°¯\©&¸Œ¬f¶A¯\°¨d¶Ã¯\Ê®P¨Œ¼4¯\¬´>¸ŒW­ S¨QÉ
Ú®²­³•Éʬ´LÊð&¬ªT¼­¬¹Á ¬¯\°¬f¶A¯&®²©&¼4©&®…¯\¸¯\¬¨d¶Ó¿ J  %ùfR\K+g ¬Åd®²¶¯\Ê®…°&®´…¨d¶°¯\©\¸¬f¶A¯\°…±^¯\Ê4®2­f¨Œ·¾p¼y¨d­ˆ¸©N°&®…¶°&¨d©
°&®…¼¸Œ©\¸Í¯\®²°¯\Ê4®"ªT¨Œ¯&¬f¨d¶TÚ®²­³°º´LÊÖ¯&ʸ¯N¯&Ê®´²¨dªo¼y¨d¶®²¶A¯³º4®¯&¨PªT¨¯\¬f¨Œ¶¡¸Œ­f¨d¶4·2¯\Ê®"Ú4Ì4¸¯&¬f¨d¶Ÿ³4¬f©&®…´…¯&¬f¨d¶
¨d¶­ÂŸ¸Œ¼¼y®>¸©&°¬f¶Ÿ¯\Ê4®¢©&¸Œ³¬f¸Œ(­ S¨QÉ¿ wÊ®¢©&¨Œ¯\¸¯\¬¨d¶¸ŒL­ S¨QÉ쬏°w³º®®²¶A¯\¬©&®…­Â ¯&¨’ªT¨Œ¯&¬f¨d¶Ÿ¼y®²©¼^®…¶³¬´²º­f¸Œ©¯\¨
¯\Ê®Úr̸ͯ\¬f¨Œ¶‹³¬f©®²´Ž¯\¬f¨Œ¶Ód¿ wÊVº°²±r¯\Ê®©¨Œ¯L¸Í¯\¬f¨Œ¶¸Œ­%´²¨dªo¼y¨d¶®²¶A¯w¨¹
¯&Êh® S¨QÉ ´>¸¶Ÿ«^®º4°&®²³¡³¬©&®…´…¯\­Â’¯\¨P¬f¶r¹‘®²©
¯\Ê®³¬©&®…´…¯\¬¨d¶¡¨Œ¹
¯\Ê®©&®…»Vº¬f©®²³Ö¸Œ³ÍǺ°¯\ªo®²¶A¯¯\¨’Ê®>¸³¬f¶·’³¬©&®…´…¯\¬¨d¶Ó¿ ir®…®TÛ 4 Þ)¹‘¨Œ©¸P¹‘º4­f­Ó³®…©&¬ÅŒ¸Í¯\¬f¨Œ¶Ó¿
á}¨QÉ!®…Åd®…©²±y¯\Ê®…©&®¢¬f°¸’Ú¶¸Œ­¸Œª«4¬f·dº¬¯pÂyñ<¹‘¨d©w¯\Ê®°\¸ªT®Ê®²¸Œ³¬f¶4·T³¬©&®…´…¯\¬¨d¶Ó±4É!¨d©­f³¼^¨d¬¶A¯\°w¹‘º©¯&Ê®²©w¹‘©¨dª
¯\Ê® ©&¨d«y¨Œ¯2¯&ʸŒ¶¯&Ê® Ú4Ì4¸¯&¬f¨d¶¼y¨d¬¶V¯2©&®…°&º­¯¬f¶ß©&¨Œ¯\¸¯\¬¨d¶¸Œ­ S¨QÉ'¬f¶ß¯&Ê®‹¨d¼4¼^¨d°¬¯&® ³¬©&®²´Ž¯\¬¨d¶ß¯\ʸŒ¶¹‘¨d©
¼^¨Œ¬f¶A¯\°¢¯\ʸ¯2¸Œ©® ´²­¨d°&®…©²j¿ wʬf°¢´>¸¶«^®T©&®…°&¨d­Åd®²³”º4°&¬f¶4·•°&¬¯\º¸Í¯\¬f¨Œ¶¸Œ­3¸>ÉN¸©&®²¶4®²°&°P¨¹N¯\Ê® ®²¶AÅJ¬f©&¨Œ¶ªT®…¶A¯>¿
&
 Right-of
Close to (turn left)
(stop) Not close to
(move forward)

Left-of
(turn left)

<>=k? <GA.?

E§ ¬f·Œº©&® ñêí¸Jë
wÊ®¨d¼4¯&¬f´>¸­7S¨QÉ+«¸Œ°&®…³T³4¨r´LµV¬f¶4·"ªT®Ž¯\ʨJ³PÉN¸Œ°<¯&®²°¯\®²³o¬f¶o¸°¬fªº­f¸¯\¨Œ©²±Œ¬¶¯\ʬ°<®ŽÌ4¸Œªo¼­®Œ±

¬¯&°P¬f¶¬¯\¬ˆ¸­-Ê4®>¸Œ³¬¶·ÃɸŒ° 4 ܕ³®²·Œ©&®²®…°o¯&¨Ã¯\Ê®Ÿ³4¬f©&®…´…¯&¬f¨d¶”¨Œ¹¯\Ê4®Ÿ¨d«JÇ®²´Ž¯>¿lwÊ4¬f°P¼­¨Œ¯P°&Ê4¨GÉ°P¯&Ê®¡©&¨d«y¨Œ¯
ªT¨QÅJ¬f¶4·“¨d¶’¯\Ê4®}·d©¨dº¶³o¼­f¸Œ¶®Œ±J¸Œ¶4³ ¬¯&°3³¬f©®²´…¯&¬f¨d¶’´…¨d¶AÅd®²©·d¬f¶4·“¯\¨¢«y®}Ê®²¸Œ³¬¶·2¸¼¼©&¨>Ìr¬fªT¸¯&®²­Â“¯\¨QɸŒ©&³°
¯\Ê®¨d«VÇ®²´…¯2êz¸Í¯w¯\Ê®¨d©¬f·d¬¶^ëL¿êí«^ë!Ô´Ž¯\¬¨d¶¡´>¸¯&®²·d¨d©¬f®…°N¹‘¨d©N¯\ʮǨd¬¶A¯¸Œ¶·d­®«¸Œ°®²³•³4¨r´LµV¬f¶4·o°ÂJ°¯\®…ªÖ¿
§¨d©®ŽÌ4¸Œªo¼­f®±E¬¹-¯\Ê4®T¯L¸©&·d®Ž¯“¬°“¨Œ¶È¯\Ê4P® S¨J¨d©…±
¯&Ê®²¶”¸­f­v¼y¨d¬f¶A¯\°«y®²­¨GÉ ¬¯¢¬f¶Æ¯\Ê®o¬fªT¸Œ·d®o¸Œ©®’´²­f¨Œ°&®²©¯\¨
¯\Ê®’©¨d«^¨¯¯\ʸ¶Æ¯&Ê®P¯L¸Œ©·d®…¯²¿mwÊVº°…±
¯&Ê®P©&¨d«y¨Œ¯"´²¸Œ¶Ë®²°¯&¬fªT¸¯\®2Ê®>¸Œ³4¬f¶·Ÿ³¬©&®…´…¯\¬¨d¶Ä«¸Œ°®²³Ë¨d¶Ä¯\Ê®’°¬f·d¶
¨Œ¹E¯&Ê®©&¨Œ¯\¸¯\¬¨d¶¸Œ(­ S¨Qɱ^¸¶³Ö´…¨d¶A¯\©¨d­)¬¯&°³4¬f©&®…´…¯&¬f¨d¶Ÿ¹‘¨d©w³¨J´LµV¬f¶·o¬f¶¸’´²­¨d°&®…³Ö¼y®²©´²®²¼r¯\¬f¨ŒL¶ eŒ¸Œ´Ž¯\¬f¨Œ¶­¨r¨Œ¼Ó¿
ã¨r´LµV¬¶·´>¸¶’¯&ÊVº°v«^®¼^®…©¹‘¨d©ªT®…³2ɬ¯&ʨdº4¯v´²¸Œªo®²©\¸}´>¸Œ­¬f«©&¸¯L¸¯&¬f¨d¶¢¨d©<µV¶¨QÉ­®²³·d®¨Œ¹dǨd¬f¶A¯v¸¶·d­f®…°²¿ wʬf°
¯pÂJ¼^®¢¨Œ¹E¸Œ´…¯&¬f¨d¶¡´>¸Í¯\®²·Œ¨d©&¬°\¸¯&¬f¨d¶‹¬°N¯pÂJ¼¬f´²¸Œ­)¨Œ¹E¸Œ´…¯&¬Åd®¼y®²©´²®²¼r¯\º¸Œ­°Âr°¯\®…ªT°…±jêh®Œ¿å·¿Û[ç 4 ޑëL¿
npo #y$/+  vr
Î !q  ` bcE( r  3t  s4 !& ( 30
Àp¶+Û ÞjÉ®P³4®…Åd®…­f¨d¼y®²³Ë¸‹³¨J´LµV¬f¶·Ÿ°ÂJ°¯\®…ª ¹‘¨d©­®²·d·Œ®²³Ã©¨d«^¨¯\°²¿ wʬ°"°ÂJ°¯&®²ª Úr̸ͯ\®²³Æ¨d¶Ã¯&Ê®P¨d«JÇ®…´…¯¨Œ¹
¬f¶A¯\®…©&®…°¯>±
¸Œ¶4³ËʸŒ³Ä¯pɨŸÅ¸Œ©&¬f¸Œ«­®²°¯&¨Ÿ´²¨d¶A¯\©¨d­z±)¬¯\°Ê4®>¸Œ³¬¶·Ÿ³¬©&®²´Ž¯\¬¨d¶Æ¸Œ¶³Æ¬¯&°"¸Œ¼¼4©&¨A¸Œ´LÊË°&¼y®²®…³Ó%¿ wÊ®
©&¨d«y¨Œ¯!°&ʨŒº­f³TªT¨QÅd®}©&¬·dÊA¯º¼T¯\¨“¯&Ê®}¯\¸Œ©&·Œ®…¯¸Œ¶³‹°¯\¨d¼ ÉÊ4®²¶‹¬¯·d®Ž¯\°-¯\Ê®…©&®Œ¿ J  %ùÄLõ K wÊ4®²©&®¸Œ©&®¹‘¨dº©
¸Œ´…¯&¬f¨d¶¡´²¸¯\®…·d¨d©&¬®²°…±¯\º©¶Ö­®…¹Á¯>±¨Œ©N¯\º©¶Ö©¬f·dÊA¯w¯\¨oÊ®²¸Œ³¯&¨QÉN¸Œ©³°¯\Ê®¯L¸Œ©·d®…¯²±¸Œ¶³°¯&¨d¼ꑯ\Ê®©¨d«^¨¯¬f°¸¯
¯\Ê®¯L¸Œ©·d®…¯ŽëL±J¨d©-ªT¨QÅd®¹‘¨d©É¸Œ©³ÃêíʸŒ°&¶Óï|¯Œ®…¯©&®>¸´LÊ®²³‹¯&Ê®}¯\¸Œ©&·d®Ž¯…ë\¿ J 2 %ùjM\K Ô·A¸¬f¶’¯\Ê®¯L¸Œ°µo¬¶AÅd¨d­Åd®²°
Ú4Ì4¸¯\¬¶·P¨d¶‹¯&Ê®¯L¸©&·d®Ž¯w¨d«JÇ®…´…¯>¿ wʬ°¯\¬ªT®±4ʨQÉ!®…ÅŒ®²©²±É!®ª ¸µŒ®®,Ì4¼4­f¬f´…¬¯º°&®"¨Œ¹)¯&Ê®!Ǩd¬¶V¯N¸Œ¶·d­®²°…¿<Õß®
¸Œ­f°¨ª ¸Œµ®wº°&®¨Œ¹y¯\Ê4®¹h¸´…¯v¯\ʸ¯v¯\Ê®w©¨d«y¨Œ¯>ïå°vÊ®>¸³T¬°v®²­f®ŽÅ¸¯\®…³T¸«^¨QÅd®w¯&Ê®·d©&¨dº4¶³Ó±V¸Œ¶³o¸Œ°°&ºªo®w¯\ʸͯ3¬¯
¬f°w°¨dªT®ŽÉʸ¯Ê¬·dÊ®²©N¯&ʸŒ¶¡¯\Ê®"Ú4Ì4¸¯&®²³¡¯L¸Œ©·d®…¯²±Éʬf´Lʝ¬f°¸°&°&º4ªT®…³¯&¨o­¬f®¨Œ¶¯&Ê®·d©&¨dº4¶³Ó¿ *}¨¯\®"¯\ʸͯ>±
«AÂó®ŽÚ¶¬¯&¬f¨d¶¦±y¯&Ê®2Ú4Ì4¸¯&®²³Ä¨d«JÇ®…´…¯­¬f®²°¨d¶•¯&Ê®¨d¼r¯\¬f´²¸Œ­j¸ÍÌr¬f°¨¹v¯\Ê®´²¸Œªo®²©\¸r¿ J 2 %ùlRNK ÀX¹3¯\Ê®2©&¨d«y¨Œ¯
¯\º©¶°“¯&¨•©®²³º´…®T¯&Ê® ´²¸ŒªT®…©\¸¡¼¸Œ¶”¸¶·d­f®P¯&1¨ Q²®…©&¨Ö¬¯É¬f­­v«^®oÊ®>¸³¬f¶·Ö³¬f©®²´Ž¯\­Â•¯\¨QÉN¸©&³°2¯&Ê®o¨d«JÇ®²´Ž¯>¿
Ô­°&¨±J¯\Ê®¯&¬f­¯¸Œ¶·Œ­f®¬°N¼©¨d¼^¨Œ©¯\¬¨d¶¸Œ­¯&¨¯\Ê®"³¬°¯\¸Œ¶´²®}¹‘©&¨dª½¯\Ê®¯\¸Œ©&·Œ®…¯êh©&®²­f¸¯\®…³‹«A ¯\¸Œv¶ uVëL¿3§¨d©!­ˆ¸Œ©·d®
³¬f°¯L¸Œ¶4´²®²°v¯&ʬf°v©®²­ˆ¸Í¯\¬f¨Œ¶2ɬf­­4¶¨Œ¯3«y®¸Œ´²´…º©\¸Í¯\®Œ±VʨQÉ!®…ÅŒ®²©²±VÉÊ®…¶T¯\Ê4®w©&¨d«y¨Œ¯3ªo¨GŌ®²°3´²­¨d°&®N¯&¨"¯\Ê®wÚ4Ì4¸¯&¬f¨d¶
¨d«JÇ®…´…¯>±¯&Ê®ªT®²¸Œ°&º©®"ɬf­f­¦¸Œ­f­¨QÉô³4¬f°&´…©&¬ªT¬¶¸¯\¬¨d¶ ¨¹<¸¯¾X¯\¸Œ©&·d®Ž¯¸Œ¶³¡¶¨Œ¯¾O¸¯¾X¯\¸Œ©&·d®Ž¯w´>¸Í¯\®²·Œ¨d©&¬®²°²¿
wtx !#0& %  y  '
 z  [+[" v#y
'(4 }4 !""â ' !0î - v(
Õß®oʸ>Åd®T¼©&®…°&®…¶V¯&®²³Ø¸¡ªT®Ž¯\ʨJ³¨d­¨d·ŒÂÖ¹‘¨d©³®…ÅŒ®²­f¨Œ¼¬f¶·®²ª2«^¨J³¬f®…³ËÅJ¬°&¬f¨Œ¶r¾p·dº4¬f³®…³Ä°ÂJ°¯&®²ªo°“¹‘¨d©ªo¨d«¬f­®
©&¨d«y¨Œ¯&°²P¿ wʬf°ªo®…¯\Ê4¨r³ÆʸŒ°«y®²®…¶ßº°®²³È¯&¨Ö³4®…Åd®…­f¨d¼Æ¯\Ê4©&®²®o³¨J´LµV¬f¶·¡°ÂJ°¯\®²ªo°²¿TÔ­¯\ʨdº4·dÊ߸­f­E¨¹¯\Ê4®²°&®
¸Œ¼¼­¬f´²¸¯\¬¨d¶°¼y®²©¯L¸Œ¬¶‹¯\¨P³¨J´LµV¬f¶·4±4¯\ʬ°N¬f°N¸’·Œ®²¶®…©\¸Œ­Óªo®…¯&ʨJ³¨d­f¨Œ·ŒÂ’¹‘¨d©N³®ŽÅd®²­¨d¼¬¶·’®²ª«y¨r³4¬f®²³TÅr¬°&¬¨d¶r¾
·dº¬³®²³ªT¨Œ«¬f­®©&¨Œ«^¨Œ¯°ÂJ°¯\®…ªT°…¿
Àp¶î®>¸´LÊð°ÂJ°¯&®²ª±¼4©&¨d«­®²ªo°o¯&ʸ¯‹¸Œ©&®¶¨Œ¯ °&¨d­ÅŒ¸«­f®±N¨d©o´²¨dªo¼º4¯\¸¯\¬¨d¶¸Œ­­Â®,Ìr¼^®…¶°&¬Åd®•¯&¨°&¨d­Åd®±¸Œ©&®
ʸŒ¶³4­f®²³®…Ño´…¬f®²¶A¯&­ÂŸº°¬f¶·o¸Œ¶®²ª2«^¨J³¬f®…³¡¸Œ¼¼©¨A¸Œ´LÊÓ¿ wÊ®"Ú4Ì4¸¯&¬f¨d¶r¾X«¸Œ°®²³ªT®Ž¯\ʨJ³°w°&¬ªT¼4­f¬¹ÁÂo³¨J´LµV¬f¶·
4
 «AÂ+©®²ªo¨QÅr¬¶·Æ¯\Ê®©&®…»Jº4¬f©&®…ªT®…¶A¯T¹‘¨d©’®,Ì4¼4­f¬f´…¬¯’¼y¨d°&®Ö®²°¯\¬fªT¸¯&¬f¨d¶Ó±-¸Œ¶4³+É!¨d©&­³+´…¨r¨d©³¬f¶¸¯\®…°²¿{wÊ©¨dº·dÊ
®²ª«y¨J³¬fªo®²¶A¯>±yÉ®o¸Œ©®P¸Œ«­®2¯\¨ ´²¨dº¼4­f®P¼y®²©&´…®²¼4¯&¬f¨d¶Æ¸¶³Æ¸Œ´…¯&¬f¨d¶Ä³4¬f©&®…´…¯&­Âd¿ÀX¯"¬°É!®²­f­8¾pµV¶¨Qɶï\ʸ¯"¬¯¬f°
´²¨dªo¼­®ŽÌ¯\¨¬f¶A¯\®…©&¼©®…¯<Ê4®>¸Œ³¬¶·³4¬f©&®…´…¯&¬f¨d¶¢«¸Œ°®²³P¨d¶­Â¨d¶2¨Œ¼4¯\¬´>¸Œ|­ S¨QÉجf¶r¹‘¨d©&ªT¸¯\¬¨d¶TÛ8ç ±Vç & Þí¿<á}¨QÉ!®…ÅŒ®²©²±
ÉÊ®²¶o¯\Ê®¨d¼r¯\¬f´²¸Œ7­ S¨GÉ׬f¶4¹‘¨Œ©&ªT¸¯\¬¨d¶¸Œ©¬f°®²°v¹‘©&¨dª¸"©&¨d«y¨Œ¯v¯\ʸ¯3¬f°v¼º4©&¼y¨d°&®Ž¹‘º­f­Â2Ú4Ì4¸¯&¬f¶·"¨d¶’¯\Ê4®N¯L¸Œ©·d®…¯
¨d«JÇ®…´…¯>±rªT¨QÅJ¬f¶4·“¨d¶o¯\Ê®·d©¨dº¶³r¾X¼­ˆ¸¶®Œ±J¸Œ¶³TµV¶¨QÉ°N°&¨dªo®¬f¶4¹‘¨d©ª ¸Í¯\¬f¨Œ¶P¸Œ«y¨dº4¯!¬¯\°3®…¶VÅJ¬©&¨d¶ªo®²¶A¯¬¯-´>¸Œ¶
®ŽÌr¼­¨d¬¯"¯&Ê®²°® ®²ª2«^¨J³¬f®…³Ä¹h¸Œ´…¯&¨d©&°¢¸Œ¶³Ë¬f°¢¸Œ«4­f®¯\¨Ö¸Œ´Ž¯“°º´²´…®²°°¹‘º­­Âd¿ §E¬f¶¸Œ­f­Âd± & 㽨d«JÇ®…´…¯“©&®…´²¨d·d¶4¬¯\¬¨d¶
¹‘©&¨dª ¸Œ¶AÂßÅJ¬f®ŽÉ⭏®>¸Œ³4°2¯\¨Ä°®>¸Œ©´LÊ+¼©¨d«­®²ªo°¡9Û }GÞNɬ¯&ʔ­ˆ¸Œ©·d® ¶Vºª2«^®…©&°’¨Œ¹ÅJ¬f®…É°P¨Œ©´²¨dªo¼º4¯\¸¯\¬¨d¶¸Œ­­Â
¬f¶A¯\®…¶°&¬Åd® & ã ªo¨r³4®²­f°…¿2ó-ÂîŽÌr¼­¬f´…¬¯\­Â´²¨d¶°¬f³®…©&¬¶·Ÿ®²ª«y¨J³¬fªo®²¶A¯>±y¹h¸Œ©ªo¨d©&®2®…Ño´²¬®²¶A¯¢ªT¨J³®²­°´²¸Œ¶Æ«^®
³®²©¬ÅŒ®²³Ó¿
~r €+‚+ƒ…„I+†
‡ t‰ˆ x C u!=O9Œ6?Ÿ;Q?VR xôx S I = H U nQb u IAKpD ? DGF"DGI B?V;ŽmŒ68ZG; K 6 D ?68?Ÿ; F ;Q? IAe ;Q7 KOI =6?AZP@,?Ímd6[= D ? F @,? KLb‹ŠŒŒ
Œ
/Ž2|‘H’
“2•”“—–D“˜š™š›œ‘h27+ž Ÿ|˜$“2¡˜š™)“2 £` ¢¤š¥¦O§™ ¥¥3¨ª©3¥3¨@¥A`yu I Z I B K t¬««>s b
‡§­ ˆm®.:<;>9Q@L=;Q?VRT5î:E;>=?A@,B b u!?T68? K @LZQ=; K @,Ro;Q7 K 6[mQ@ H @L=7\@ HVKOI ;QY_@,UV;ŽmŒ6 DGI = eD = D _dŠp@,7 K 68? K @L=;Q7 K 6 D ? b k @,7Ud
?J687L;QY x @ HD = K ­ sQsAªt ¯ ­ qŒ`l?V6[mQ@L=68B Khn¢DQe ST@,Y[_ DGI =?A@Q`dMN687>`Jq>sAtLsd`ru I B K =;QY868;d` ­ sQsAt b
‡ kq ˆ¢5ô:<;>=?A@,Bw;Q?V° R ®cd;Q?JRV68?V6 b u7 K 6[mQ@R D 7O9d68?AZ_J;QBp@,R D ? K UA@= DQK ; K 6 D ?V;QY)7 DGF"HD ?A@,? KwDQe Y D Z> HD Y8;>= DQHVK 687
˜JDŽgb WX? ž 0 ± k BO;Q6d;Q?V R ±30 ²w]y@L@Q`G@,RV6 KpD =B,` ž+³³µ´%¶Ž“ª›D·¸·¸2™¹.º2‘“»˜¹¼|·ž‘D™)2½³\“2¬»¬·HŽ¾·H7›‰·“¿³\“2¡À7Ÿ|˜0·œŽ
´(™C‘D™)“2 ` H ;>ZQ@,  B «¨¨ª©Á«¢>sd` ­ sQsQs b
‡ ¥2ˆ¢5Ä:<;>=?V@,B^;Q?V R ®ÈcA;Q?VRV68?V6 b aw6[=@,7 K 6 D ?7 D ? K = D Y eD =Ó;Q?;Q7 K 68mQ@<R D 79Œ68?VZ3_@,UV;,md6 DGI =%_r;QBp@,R D ? K UA@j= DQK ; K 6 D ?V;QY
7 DGFH4D ?V@,? KEDQe Y D Z> HD Y8;>= DQHVK 687 ˜VDŽgb WX? Œ ŸŽ“DÀ/·¾2…³\“2¬»¬·HŽ¾·H7›‰·h“21³\“2¡À7Ÿ|˜0·œŽ%´(™C‘D™)“2PÂÁÃ3×à `Vm D Y IJF @ ­ `
H ;>ZQ@,B¬t ¢—Ĭ©r¬t Ådt>` ­ sQsQs b
‡ ¨kˆ¢5ÈS :<;>=?V@,B;Q?VÇ R Æ^Èß]^6 Ib i F _ D RJ6[@,R}7 DGFHJIAK @L=md68BO6 D ? efD = F"D _J6Y[@j= D _ DQK B b Wz? Š ³ Š ¶É#Ê ËÌIÍ ŠŒ
ŒŒ#Š 
˜$’
³\“¬»H’“2 Š 
˜0·HιÎϙк.·œ
˜É
™Ñº—72ÎN¶Ž“ª›D·‰‘D‘D™¹.º \` Ò
= D 7\@L@,RJ68?AZGB DQe<K UA@o¬t ««—ÄPWziji
i+Wz? K @L=?V; K 6 D ?V;QY CED ? e @L=O@,?V7\@
D ?’Wz? K @,Y8Y86[ZQ@,? K Ò= D 7\@,BOB68?AZ¢c n B K @ F BL` H ;>ZQ@,B}t,q «¨ª©rt,q ««Œ`%5-@ g ²G@L=BO@ n `JlNcAu7` ÓN7 KLb ¬t ««—Ä b WXi
i
i b
‡ ¢kˆ¢5 Sæ:<;>=?A@,Bw;Q?V½ R Æ#È ]^6 Ib M68B6 D ?oZ I 68RA@,R 7L6[=7 IVF ?J;,md6[ZG; K 68?AZ2; IAKpD ? DGF"DGI B= D _ DQK B vb Š 
˜$’Õԁ“2ŸŽDÖ2ÎW“»
¶˜)˜0·HŽ‰P”·¾›D“¾º
™¹˜š™)“2;2Ö ×j—›H¼.™CÖ· Š 7˜0·HιÎϙк.·œ7›D· `%Ht ¥/¤$¢.¦O9™ ¢Å«ª©ÄŒHt ¥A` ­ sQsQs b
‡ ĪˆÇ²2a IAH = @ Ø b Ù¼|·2™C‘H“2Ž3·œŽÇ“»+˜¹¼3™CÁº2‘Úh¡·¬˜$DÀ
¼3Û@‘‰™š›D2Î|»H“ŸÁÖ—˜š™)“2|‘‹“» ˜¹¼·‹2™>‘‰Ÿ/™¹˜šÛ^“»h‘œ›œ™$·H7›‰· b ±!;>=Om>;>=R
l?V6[mQ@L=BO6 Khn Ò=O@,BBL` C ; F _V=68RAZQ@Q`VS ;QBOBL`^¬t ««Qq b
‡ ÅkˆmÓÝÜJ; I ZQ@L=;QBL/` ²S I ?JR n `Œ5”uU I Šp;d` C a n @L=,`AÝ u Ò¦@,? K Y8;Q?JR` x ²;Q68?`d;Q?V‹R Þ+Wz9Q@ I 7UV6 b%ž U n ;QB H @,7 K ZQ=; H UVB
;>=@-? DQK ¤ n @ K ¦ H =;Q7 K 687L;QY efD =j7 DGFHJIAK @L=md68BO6 D ? b Wzà? ß “2ŽD᪑¸¼“DÀ¿“2‹âv™¹Ž·¸›¬˜š™)“2|‘™¹‹ž Ÿ|˜$“2¡˜0·¾ã³Lž âc͚䠏@‘¬·¾
´(™C‘D™)“2 ` H ;>ZQ@,  B «—Ĭ©rtL2s ¥A`¦¬t ««dt b
W ± D =B g 68Y8Y b MN68B I ;QY“7 D Y8Y868B6 D ? ;Žm D 68RV;Q?V7\@_ n Bp@LZ F @,? K ; K 6 D ? b Wz? Š ”cådÉæÊ Ë¬ç
’‹¶dŽ“¬›‰·¸·¾™C.º‘l“»˜¹¼·
‡ «kˆ°
ŠŒŒ
ŒNè ” É
Ô è"éWŠÇŠ 7˜0·HŽ‰7˜š™)“272Î ³\“2¬»¬·HŽ¾·H7›‰·¿“2 Š 7˜0·HιÎϙк.·œ
˜ ”“3–¸“˜¹‘ꏐ7;É/Û@‘‰˜0·H¡h‘œ’žë@27›‰·¾½”“—–D“˜š™)›
É
Ûk‘H˜0·H¡h‘c27%˜¹¼|·”·¾2Î ß “2Ž‰Îѝ ` H ;>ZQ@, B «>s ­ ©Á«>—s «Œ`)¬t ««@¥ b
‡ tL@s ˆ“z S ² D UJ?VB D ? b Ù¼·䠓¬ۋ™¹ê˜¹¼|·׿™¹7 b l?V6[mQ@L=B6 Khn“DQeC UV687L;>Z D Ò=@,BOBL` C UV67L;>Z D `y¬t «Å—Ä b
‡ tQ‰t ˆ%®ð]y;>9 D – b ß “2¡Ç·œ|ìNí
™CŽ¾·‰ìd27âh2.ºÁ·HŽ“2ŸÁ‘mÙ¼3™CÁº2‘ b l!?V68mQ@L=BO6 Kín“DQeC UV687L;>Z D Ò=O@,BOB,™ C UV687L;>Z D `%¬t ««>s b
‡ t ­ ˆ‹± C ] D ?AZ I @ K 0 ±6[ZQZG68?JB-;Q?V^ R ÞîÒ=;  RJ? nQb k UA@6? K @L= H =O@ K ; K 6 D ? DQeF"D mŒ6?AZ=O@ K 68?J;QY%6 F ;>ZQ@,B b ¶dŽ“¬›k’(”“2ۗ2Î
É֓¬›œ™$·œ˜šÛ%“» ïL“7—“2 ` ­ —s Ō™åq Ũª©Œq «—Ä`
¬t «Å>s b
‡ t,kq ˆ‹²;± x 6[@LZQ@L=“;Q?Jð R ± k ]y; g3KpD ? b Ò
= D 7\@,BOBO68?AZŸRV6[–4@L=O@,? K 68;QYv6 F ;>ZQ@ F"DQK 6 D ? b ÔÁ“2ŸŽœ’ å(À֘š™)›¸2ÎÉ֓¬›œ™$·œ˜šÛ#“»
žµ¡·œŽD™)›¸ ž ` ™åq ¨@¥@©Œq ¢>sd`¬t «Å¨ b
­
‡ Ht ¥2ˆ‹²¢cA;Q? KpD BX‘MN687 KpD =;Q?VP R ®ðcd;Q?VRV6?V6 b i F _@,RVRA@,R2md68B I ;QYr_@,UV;,md6 DGI =B efD =3?V;ŽmŒ6[ZG; K 6 D ? b ”“—–¸“—˜š™š›œ‘c27+ž Ÿ|˜$“Í
¡“2ŸÁ‘IÉ/Û@‘H˜0·H¡h‘ `y¬t «¤‘q…$ ¥¦O™ ­ ««ª©Œqdt,qA`yST;>=7U¬t ««—Ä b