Academic Course: 06 Morphogenetic Engineering

Designed by René Doursat
Morphogenetic Engineering

(ECSO)1. Engineering & Control of Self-Organization
(ME)2. Morphogenetic Engineering
i.e. “Architectures without Architects”
(EE)3. Example: Embryomorphic Engineering
i.e. “Artificial Life Evo-Devo”
3.a. MapDevo
Modular Architecture by Programmable Development
3.b. ProgLim
Program-Limited Aggregation
MORPHOGENETIC ENGINEERING

 Between natural and engineered emergence
CS (ICT) Engineering: creating and programming
a new, artificial self-organization / emergence
 (Complex) Multi-Agent Systems (MAS)
CS Science: observing and understanding "natural",
spontaneous emergence (including human-caused)
 Agent-Based Modeling (ABM)
Engineering & Control of Self-Organization:
fostering and guiding complex systems
at the level of their elements
1. Engineering & Control of Self-Organization

 Exporting models of natural CS to ICT: “(bio-)inspiration”
 already a tradition...
ex: genes & evolution
laws of genetics
genetic program,
binary code, mutation
genetic algorithms (GAs),
& evolutionary computation
for search & optimization
ex: neurons & brain
biological neural models
binary neuron,
linear synapse
artificial neural networks
(ANNs) applied to machine
learning & classification
ex: ant colonies / bird flocks
trails, swarms / collective motion
move, deposit, follow “pheromone” /
separation, alignment, cohesion (“boids”)
ant colony optimization (ACO)
graph theoretic & networking problems /
particle swarm optimization (PSO)
“flying over” solutions in high-D spaces
TODAY: simulated in a Turing machine / von Neumann architecture

 ... and looping back onto an unconventional physical implementation
to fully exploit the "in materio" computational efficiency
DNA computing
synthetic biology
chemical, wave-based
computing
TOMORROW: running in truly parallel roboware, bioware, nanoware, etc.
swarm robotics

specific natural or societal
complex system
model simulating this system
generic principles and mechanisms
(schematization, caricature)
new computational discipline
exploiting these principles
to solve ICT problems
 common shortcoming: the classical "(over-
)engineering" attitude
 reintroducing too much exogenous, top-
down design/control
 exploit and only “influence” the natural
endogenous properties!
 breaking things apart too much
 keep the system internally complex
and self-organized!
 the stepwise fallacy: start with simple
tasks (ex: XOR) and build up from there
 go for the collective attractors right
away! (Kauffman’s “order for free”)
 keeping self-organization, but forcing it
into rigid design (logic gate flow  RBN)
 clinging to computing universality
 specialization is more promising!

John Holland, founder of GAs and major
promoter of Complex (Adaptive) Systems: his
titles  ~100% CAS
• Hidden order: How adaptation builds complexity
• Emergence: From chaos to order
• Artificial adaptive agents in economic theory
• Outline for a logical theory of adaptive systems
• Studying complex adaptive systems
• Exploring the evolution of complexity in signaling
networks
• Can there be a unified theory of CAS?
• etc.
GECCO, among the biggest (and most
selective) EC conferences:
its tracks in 2009  ~15% CAS?
(... individuals internally complex?)
• ACO and Swarm Intelligence
• Artificial Life, Evolutionary Robotics, Adaptive
Behavior, Evolvable Hardware
• Bioinformatics and Computational Biology
• Combinatorial Optimization and Metaheuristics
• Estimation of Distribution Algorithms
• Evolution Strategies and Evolutionary Programming
• Evolutionary Multiobjective Optimization
• Generative and Developmental Systems
• Genetic Algorithms
• Genetic Programming
• Genetics-Based Machine Learning
• Parallel Evolutionary Systems
• Real World Application
• Search Based Software Engineering
 Ex. of “over-engineered” bio-inspired domain: evo computation

 Other examples of “over-engineered” bio-inspired domains:
 artificial neural networks (IJCNN, etc.)
 multi-agent systems (AAMAS, etc.)
 ... even artificial life! (Alife, ECAL) is playing down the self-
organization and complex systems properties of living matter
 Conversely: “natural” complex systems conferences don’t show
much concern for design, engineering or control issues
 overcrowded with (statistical) physicists (ICCS, ECCS, etc.)
 or overcrowded with biologists

Designed by René Doursat http://iscpif.fr/ecso2014
... and that’s why we need

2. Morphogenetic Engineering
(ME)

the brain organisms ant trails
termite
mounds
animal
flocks
cities,
populations
social networks
markets,
economy
Internet,
Web
physical
patterns
living cell
biological
patterns
animals
humans
& tech
molecules
cells
 All agent types: molecules, cells, animals, humans & tech
2. Architectures Without Architects

 ... yet, even human-caused systems are
"natural" in the sense of their
unplanned, spontaneous emergence
the brain organisms ant trails
termite
mounds
animal
flocks
physical
patterns
living cell
biological
patterns
 biology strikingly demonstrates
the possibility of combining
pure self-organization and
elaborate architecture, i.e.:
 a non-trivial, sophisticated morphology
 hierarchical (multi-scale): regions, parts, details
 modular: reuse of parts, quasi-repetition
 heterogeneous: differentiation, division of labor
 random at agent level, reproducible at system level
 "Simple"/random vs. architectured complex systems
2. Architectures Without Architects

Doursat, Sayama & Michel (2012)

 ME brings a new focus inside the ECSO family
 exploring the artificial design and implementation of decentralized
systems capable of developing elaborate, heterogeneous
morphologies without central planning or external lead
Morphogenetic Engineering (ME) is about designing the
agents of self-organized architectures... not the architectures directly
 swarm robotics,
modular/reconfigurable robotics
 mobile ad hoc networks,
sensor-actuator networks
 synthetic biology, etc.
 Related emerging ICT disciplines and application domains
 amorphous/spatial computing (MIT, Fr.)
 organic computing (DFG, Germany)
 pervasive adaptation (FET, EU)
 ubiquitous computing (PARC)
 programmable matter (CMU)

http://iscpif.fr/MEW2009
1st Morphogenetic Engineering Workshop, ISC, Paris 2009
http://iridia.ulb.ac.be/ants2010
2nd Morphogenetic Engineering Session, ANTS 2010, Brussels
Morphogenetic Engineering:
Toward Programmable Complex Systems
Fall 2012, SpringerR. Doursat, H. Sayama & O. Michel, eds.
http://ecal11.org/workshops#mew
3rd Morphogenetic Engineering Workshop, ECAL 2011, Paris

1) O'Grady, Christensen & Dorigo
2) Jin & Meng
3) Liu & Winfield
4) Werfel
5) Arbuckle & Requicha
6) Bhalla & Bentley
7) Sayama
8) Bai & Breen
9) Nembrini & Winfield
10) Doursat, Sanchez, Dordea, Fourquet &
Kowaliw
11) Beal
12) Kowaliw & Banzhaf
13) Cussat-Blanc, Pascalie, Mazac, Luga &
Duthen
14) Montagna & Viroli
15) Michel, Spicher & Giavitto
16) Lobo, Fernandez & Vico
17) von Mammen, Phillips, Davison,
Jamniczky, Hallgrimsson & Jacob
18) Verdenal, Combes & Escobar-Gutierrez

Part I. Constructing
Relatively few robots or components
(possibly originating from a larger,
ambient pool) attach to each other or
assemble blocks together, creating
relatively precise formations. The built
structures are ``coarse-grained'', and
most of them stick figures, i.e., made
of 1-unit wide segments. The space is
the 2D plane, with occasional vertical
elevation into 3D by stacking or
folding.
Part II. Coalescing
A great number of mobile agents
exhibit heterogeneous flocking
behavior. Without literally attaching,
they aggregate and stay near each
other while moving to maintain
neighbor-to-neighbor communication
(e.g., of visual, chemical, or wireless
type). Together, they tend to form
dense, fine-grained clusters that
assume certain ``fluid'' yet stable
shapes. Most contributions focus on
simulated systems, as large-scale
robotic swarms are still too costly and
programmable flocking nano-particles
unheard of.

Part III. Developing
Closer to cell-based models of biological
morphogenesis. Starting from a single
agent and growing to a large size by
repeated, yet differential, division or
aggregation. Mechanisms underlying this
development involve one or several
biological features such as gene
regulation, molecular signalling and
chemotaxis. As a consequence, the
resulting structures exhibit properties of
biotic patterns or tissues, such as
vascularization and segmentation, or
entire organisms, such as arthropods or
branched creatures. Potential
applications range from synthetic biology
and collective robotics to computer
networks
Part IV. Generating
Morphogenetic systems are generated by
successive transformations of
components in 3D space, based on
``rewrite'' rules. These rules are formally
expressed as ``grammars'', which can be
designed by hand or evolved. The
resulting architectures have potential
applications as diverse as natural
computing, robotics, computer graphics
or plant biology.

Other, less distinguishing
features:
Robotic models or applications,
virtual or physical.
Actual physical implementations
with robots or mechanical
components.
Emphasis on various patterns,
textures, or symmetrical shapes,
rather than complicated
morphologies.
Biological models based on real
data, such as fruit fly and rye
grass.

i.e. “An Artificial Life Evo-Devo”

 already a tradition (although too re-engineered and "decomplexified")
ex: genes & evolution
laws of genetics
genetic program,
binary code, mutation
ex: neurons & brain
biological neural models
binary neuron,
linear synapse
ex: ant colonies / bird flocks
trails, swarms / collective motion
move, deposit, follow “pheromone” /
separation, alignment, cohesion (“boids”)
... or embedded in bioware, nanoware...whether simulated in a Turing machine...
3. Embryomorphic Engineering: Alife Evo-Devo

 A new line of bio-inspiration: biological morphogenesis
 designing multi-agent models for decentralized systems engineering
Doursat (2006) Doursat, Sanchez, Fernandez
Kowaliw & Vico (2013)
Doursat & Ulieru (2009) Doursat, Fourquet,
Dordea & Kowaliw (2013)
Doursat (2008, 2009)
Embryomorphic Engineering
... or embedded in bioware, nanoware...whether simulated in a Turing machine...

evolution(My)
development(m/y)
Doursat (2008, 2009)
one individual is
internally complex
learning

Nothing in biology makes sense
except in the light of evolution
—Dobzhansky, 1973
Nothing in development makes sense
except in the light of complex systems
Not much in evolution makes sense except
in the light of multicellular development
(or molecular self-assembly for unicellular organisms)
Evo-Devo 3. Embryomorphic Engineering: Alife Evo-Devo

 Development: the missing link of the Modern Synthesis...
Purves et al., Life: The Science of Biology
evolutionmutation
“When Charles Darwin proposed his theory of evolution by variation and
selection, explaining selection was his great achievement. He could not
explain variation. That was Darwin’s dilemma.”
—Marc W. Kirschner and John C. Gerhart (2005)
The Plausibility of Life, p. ix
“To understand novelty in evolution, we need to understand
organisms down to their individual building blocks, down to their
deepest components, for these are what undergo change.”
?? ??

NathanSawaya
www.brickartist.com
“To understand novelty in evolution, we need to
understand organisms down to their individual building
blocks, down to their deepest components, for these are
what undergo change.”
AmyL.Rawson
www.thirdroar.com
macroscopic,
emergent level
microscopic,
componential
level
emergence

NathanSawaya
www.brickartist.com
AmyL.Rawson
www.thirdroar.com
generic elementary
rules of self-assembly
macroscopic,
emergent level
microscopic,
componential
level
Genotype Phenotype“Transformation”?
more or less direct
representation
 ( )
emergence

3.a. MapDevo

Doursat, Sanchez, Fernandez, Kowaliw & Vico (2013)
 3D Development
3.a. MapDevo – Modular Architecture

 Body

 Limbs

Capturing the essence of morphogenesis in an Artificial Life agent model
grad1
div1
patt1
div2
grad2
patt2
div3
grad3
 Alternation of self-
positioning (div)
and self-
identifying
(grad/patt)
genotype
each agent
follows the same set
of self-architecting rules (the "genotype")
but reacts differently depending on its neighbors
patt
3
...
Doursat (2008, 2009)



pA
B
V
rr0rerc
GSA: rc < re = 1 << r0
p = 0.05
div

38
grad
EW
S
N
EW
WE WE
NS

39
I4 I6
B4
B3
patt
X Y
. . . I3 I4 I5 . . .
B1 B2 B4B3
wix,iy
GPF : {w }
wki
WE NS
Bi = (Li(X, Y)) = (wix X + wiy Y  i) Ik = i |w'ki|(w'kiBi + (1w'ki)/2)

40
I9
I1
(a) (b)
(c)
. . . . . .
WE = X NS = Y
B1 B2 B3 B4
I3 I4 I5
X Y
. . . I3 I4 I5 . . .
B1 B2 B4B3
wiX,Y
GPF
wki
 Programmed patterning (patt): the hidden embryo atlas
a) same swarm in different colormaps to visualize the agents’ internal
patterning variables X, Y, Bi and Ik (virtual in situ hybridization)
b) consolidated view of all identity regions Ik for k = 1...9
c) gene regulatory network used by each agent to calculate its expression
levels, here: B1 = (1/3  X), B3 = (2/3  Y), I4 = B1B3(1  B4), etc.

rc = .8, re = 1, r0 = 
r'e= r'0=1, p =.01
GSA
I4 I6
E(4)
W(6)
I5I4
I1
N(4)
S(4)
W(4) E(4)
SA
PF
SA4
PF4
SA6
PF6
SA
PF
SA4
PF4
SA6
PF6
all cells have same GRN, but execute different
expression paths  determination /
differentiation
microscopic (cell) randomness, but
mesoscopic (region) predictability
Doursat
(2008, 2009)

 Bones & muscles: structural differentiation and properties
(a) (b) (c)
(c)

 Locomotion and behavior by muscle contraction

 Quantitative mutations: limb thickness
GPF
GSA
33
1, 1
p = .05
g = 15
4 6
disc
GPF
GSA
11
tip p’= .05
g’= 15
GPF
GSA
11
tip p’= .05
g’= 15
GPF
GSA
33
2, 1 4 6
disc
p = .05
g = 15
GPF
GSA
11
tip p’= .05
g’= 15
GPF
GSA
33
0.5, 1 4 6
disc
p = .05
g = 15
(a) (b) (c)
wild type thin-limb thick-limb
body plan
module
limb
module
4 6
Doursat (2009)

(a) (b) (c)
antennapedia duplication
(three-limb)
divergence
(short & long-limb)
PF
SA
11
tip p’= .05
GPF
GSA
33
p = .05
4 2
disc
6
PF
SA
11
tip p’= .1
PF
SA
11
tip p’= .03
GPF
GSA
33
p = .05
4 2
disc
6
GPF
GSA
11
p’= .05tip
GPF
GSA
33
p = .05
4 2
disc
GPF
GSA
11
p’= .05tip
4
2
6
 Qualitative mutations: limb position and differentiation
antennapedia homology by duplication divergence of the homology
Doursat (2009)

production
of structural
innovation
Changing the agents’ self-architecting rules through evolution
by tinkering with the genotype, new architectures (phenotypes) can be obtained Doursat (2009)

 A simple challenge: path-based fitness
f = | end – start | / path length < 1

optimal body size? optimal limb size?

3.a. MapDevo
3.b. ProgLim
Program-Limited Aggregation

Doursat, Fourquet, Dordea & Kowaliw (2013)
3.b. ProgLim – Program-Limited Aggregation

 Preferential (“scale-free”)
Programmed
Attachment
Networking

 Preferential
Programmed
Attachment
Networking
(ProgNet)
 Diffusion-
Program-
Limited
Aggregation
(ProgLim)

 Stereotyped
development
 Environment-
Induced
Polyphenism
evol evol

Academic Course: 06 Morphogenetic Engineering

More Related Content

Similar to Academic Course: 06 Morphogenetic Engineering (20)

More from FET AWARE project - Self Awareness in Autonomic Systems (20)

Recently uploaded (20)

Academic Course: 06 Morphogenetic Engineering