Upload
others
View
2
Download
0
Embed Size (px)
Citation preview
AM, EE141, Swarm Intelligence, W6-2
Collective Building andSelf-Assembly in Natural
and Artificial Systems
Outline
• Collective building and self-assembly in natural systems• Examples• Mechanisms• Modeling• Reverse engineering with GA
• Collective building and self-assembly in artificial systems• Passive bricks• Active mechatronic units
Collective Building andSelf-Assembly in Natural
Systems
Natural Examplesof Collective Building
© Guy Theraulaz
© Guy Theraulaz
Natural Examplesof Collective Building
© Guy Theraulaz
Natural Examplesof Collective Building
twhiteMacro structure is same, micro structure variation -- the template mechanism at work.
© Guy Theraulaz
Natural Examplesof Collective Building
© Guy Theraulaz
Natural Examplesof Collective Building
© Pascal Goetgheluck
Natural Examplesof Collective Building
twhiteAgain, very similar to structures of previous pictures. Modules that are repeated again and again.
© Pascal Goetgheluck
Natural Examplesof Collective Building
© Pascal Goetgheluck
Natural Examplesof Collective Building
© Masson
Natural Examplesof Collective Building
© Masson
Natural Examplesof Collective Building
twhiteStructures can obviously be huge -- insects can do what we do with our cities!
Natural Examplesof Collective Digging
Coordination Mechanisms forCoordination Mechanisms forCoordination Mechanisms forCoordination Mechanisms forthe Building Activitythe Building Activitythe Building Activitythe Building Activity
The plan
Environmental template
11
22
Stigmergy33
Collective BuildingMechanisms
© Scott Camazine
Collective BuildingMechanisms
The plan
Collective BuildingMechanisms
twhiteNo! There's no real plan here except a stimulus response mechanism that identifies local configurations ...
It's the regularity and modularity that's key.
Collective BuildingMechanisms
The plan
Environmental template
11
22
Stigmergy33
Environmental template
• The building plan pre-exists in the environment under the form of spatialheterogeneities.
• The social insect activity only outlines these pre-existing environmentaltemplate. Environmental changes performed by insects play a minimalrole in the building activity itself.
• There are several forms of environmental template:• gradients naturally existing in the environment (humidity, temperature …)
• chemical gradients generated by one or more individuals of the colony
Collective BuildingMechanisms
Eggs
Larvae
Nest Structure in the Ant Acantholepis Custodiens
3.00 a.m. 3.00 p.m.
Pupae
T
twhiteTemplate mechanism guides the stigmergic building mechanisms -- essentially this is just the interaction of two signals.
Convective Air Flows andComplex Chemical Templates
twhiteGradients can be temperature and chemical ... and other combinations too.
© Guy Theraulaz
Convective Air Flows andComplex Chemical Templates
The plan
Environmental template
11
22
Stigmergy33
Collective BuildingMechanisms
• It defines a class of mechanisms exploited by social insects to coordinateand control their activity via indirect interactions.
• Stigmergic mechanisms can be classified in two different categories:quantitative (or continuous) stigmergy and qualitative (or discrete)stigmergy
Stimulus
Response
S1
R1
S2
R2
S3
R3
time
S 4
R4
S 5
R5
Stop
Definition
Stigmergy
Positive feedback and self-organization
Sequence of stimuli and answers quantitativelydifferent
11
Sequence of stimuli and answers qualitativelydifferent
22
The role of the two different stigmergic mechanisms in collective building
Stigmergy
• The successive stimuli quantitativelydifferentiate in their amplitude and merelymodify the answer probability of otherindividuals
• Examples :mass recruitment in antsdead ant aggregation in ant cemeterypillar building in termites
Features
Quantitative Stigmergy
Pillar building in termites
• Termites impregnate with pheromones the building material• Pheromones diffuse in the environment• Individuals carrying ground bullets follow the chemical gradient; they
climb towards the highest pheromonal concentration and drop there theirbullets
• Material drop rate is proportional to the number of active insects in thelocal region (positive feedback)
Spatial distribution of insects and their building activity are locally controlledby the pheromone density.
Quantitative Stigmergy
twhitePatterns of pillars emerge from an (initially) random distribution of materials.
Now lets add another force ...
twhiteGradient is in direction of arrows. We see that a valley appears after some time.
Sequence of stimuli and answers quantitativelydifferent
11
Sequence of stimuli and answers qualitativelydifferent
22
The role of the two different stigmergic mechamisms in collective building
Stigmergy
Self-assembly process
twhitecontinuous = pherome sensing from building materials.discrete = cell building in wasps nests
• Successive stimuli are qualitatively different. Thisprocess generates a self-assembly dynamics.
• No pheromone involved?
time
S 2
B C
•••S 3S 1
A
Features
Qualitative Stigmergy
Nest Building inNest Building inNest Building inNest Building in Polist Polist Polist Polist Wasps Wasps Wasps Wasps
© Guy Theraulaz
Nest Building in Polist Wasps
Organisation of the building activity
• It is indirectly carried out via the different, localconfigurations a wasp can find in the nest
• A probability of deposing a new cell is associatedto each configuration
S 3
S 1 S 1
S 1
S 1S 1
S 1
S 1
S 2
S 2
S 2S 2
Potential sites for building
Nest Building in Polist Wasps
Nest Building in Polist Wasps
Probability of creating a new cell given the configuration of neighboring cells
0 ,0
0 ,2
0 ,4
0 ,6
0 ,8
1 ,0
1 2 3Number of adjacent cell walls
Nest-Building ModelingNest-Building ModelingNest-Building ModelingNest-Building Modeling
Swarm on a 3D Lattice
A swarm of nest builders … agent features
• Reactive actions• Random movements on a 3D lattice, no trajectories• No embodiment (1agent = 1 single cell), no interference
(however 2 agents cannot occupy the same cell)• n agents working together means n actions (not
necessarily n dropped bricks) before next iteration starts;all n agents see the same configuration at a giveniteration
• Agents’ team is homogeneous• No global plan of the whole building• Local perception of the environment
The neighborhood of each agent is defined as the 20 cells around it (7 above, 6around, and 7 below).
Neighborhood Cell contents
X X
z + 1 z z - 1
2
22
0
01
1
2
22
01
1
2
22
0
21
1
Definition of agent ’s neighborhood for hexagonal cells
Swarm on a 3D Lattice
A
z + 1
z
z - 1
Distributed nest building: several sites are active simultaneously
Swarm on a 3D Lattice
Swarm on a 3D Lattice
Examples of building rules
z
Z+1
Z-1
• Building process is irreversible(no cell can be removed)
• Deposit rules can be strictlydeterministic or probabilistic
twhite
Example of a rule set (Polist wasps, 7 rules)
Swarm on a 3D Lattice
Figure 6.9Swarm Intelligence: Bonabeau et al
Nest-building Modeling
Obtained structures (Polists wasps, 7 rules)
With deterministic rules With probabilistic rules
Looking for Stable Nest Architectures
What kind of nest architectures can we build with the same method?
• What are the rules to implement in order to obtain stable architectures?
• An architecture is considered stable when several runs of a simulationwith the same rule set generate architectures with the same globalstructure.
• Mechanisms of stigmergic coordination reduce the number of stablearchitectures.
Agelaia (13 rules)
Examples of Stable Architectures
Parachatergus (21 rules)
Vespa (13 rules)
Stelopolybia (12 rules)
Examples of Stable Architectures
Chatergus (39 rules) Artificial Nest Structure (35 rules)
Examples of Stable Architectures
twhiteShow animations.
• The building process is carried out in successive steps. The current localconfiguration generate a stimulus different from that of the previousand of the successive configuration (qualitative stigmergy).
• Only this type of building algorithms generate coherent architectures.
• The whole set of these algorithms generates a limited number of nestshapes.
How to build a stable architecture?
Looking for Stable Nest Architectures
Reverse-Engineering: Fromthe Building to the
Individual Rules using GA
Similarities to Controller EvolutionPrey-Predator (GA, Floreano 98) Obstacle avoidance (RL, Kelly 97)
Exploration (RL, Millan 97) Foraging (GA, Jefferson 93)
Area Integration (RL,Versino 97)Exploration (RL, Hayes 01)Nest-building (Bonabeau, GA 00)
Evolutionary Encoding of theDistributed Nest Building Problem
• Phenotype: agent endowed with set of microrules• Genotype: set of microrules (one-to-one mapping with phenotype);
chromosome of variable length; 1 gene = 1 microrule.• Life span: number of iteration (e.g. 30,000 iterations, 1 iteration = all 10
agents have applied their microrule) or exhausting of max amount of bricksavailable (e.g. 500 bricks).
• Population: 80 individuals• Generations: 50• Fitness function: weighted sum of space filling and pattern replication
(arbitrary criteria based on 17 human observers who were asked to evaluatethe amount of structure in a set of 29 different patterns, [Bonabeau 2000]);
• Selection: roulette wheel• Crossover: two-points, pcrossover = 0.2• Mutation: pmutation1 = 0.9 (during life span inactive microrule);
pmutation2 = 0.01 (during life span active microrule).
Figure 6.12
Swarm Intelligence: Bonabeau et al
Evolutionary Encoding of theDistributed Nest Building Problem
• Biased evolution:• Probabilistic templates (reduction of stimulating configurations
containing a large number of bricks).• Start microrule.• No diagonal deposits (space not filled).
• Problems:• Fitness function: Functional? Esthetical? Biological plausible?
Mathematical definition?• Episthatic interactions (sequence of microrules needed for
coordinated algorithms).
twhiteStarting population is biased too: mean 50, std dev. 10
twhiteReally nice thesis on applying GP to this problem: linear representation is at fault here.
Collective Building andSelf-Assembly in Artificial
Systems
Applications• Obstacle avoidance in highly constrained and
unstructured environments• Formation of bridges, buttresses, stairs and
other structures for emergencies• Envelopment of objects
– Recovering satellites from space• Inspections in constrained environments
– Nuclear reactors• Self-organizing unfolding structures
– Space stations– Satellites– Scaffolds
Passive Bricks
Passive Bricks
• In-line 2D structures using Kheperas [Martinoli99] -> [Easton ??]
• Lionel Penrose (1898 – 1972)• Self-assembly mechanisms of genetic relevant
molecules reproduced with passive wood bricks• External energy source (shaking, human action)• Evolution and self-assembly tightly coupled• Video-tape!
twhiteTitle: Mechanical self-replication. Author(s): Lionel S. Penrose and Roger Penrose. Year(s): 1959. Model: Simple mechanical units. Implementation: Plywood. Reference(s): L. S. Penrose. ``Self-reproducing machines.'' Scientific American, Vol. 200, No. 6., pages 105-114, June 1959. Description: Lionel Penrose, aided by his son Roger Penrose, built simple mechanical units or bricks. An ensemble of such units were placed in a box, which was then shaken. As described by Penrose (1959): ``In fanciful terms, we visualized the process of mechanical self-replication proceeding somewhat as follows: Suppose we have a sack or some other container full of units jostling one another as the sack is shaken and distorted in all manner of ways. In spite of this, the units remain detached from one another. Then we put into the sack a prearranged connected structure made from units exactly similar to those already within the sack... Now we agitate the sack again in the same random and vigorous manner, with the seed structure jostling about among the neutral units. This time we find that replicas of the seed structure have been assembled from the formerly neutral or ``lifeless'' material.''
Passive BricksEvolutionary architecture: Nicolas Reeves (UQAM Montreal, Canada)• Cellular automata approach.• No genetic operator: population manager replaced by direct intervention of the
human being.
3D CAD simulations Stereolytographic sculptures
twhiteStrong resemblance to L-systems
Passive BricksEvolutionary architecture: Pablo Funes (Brandeis University, US)• GA approach, geometry- and force-based fintness functions• Off-board evolution and reproduction of results with Lego® bricks• 2D (bridge, crane) and 3D structures (table)
Set-up and objective Diagram of forces
Passive BricksEvolutionary architecture: Pablo Funes (Brandeis University, US)
2D Ex.: Bridge
Simulated bridge Real Lego bridge
twhite
Passive BricksEvolutionary architecture: Pablo Funes (Brandeis University, US)
3D Ex.: Table2D Ex.: Crane
Active Mechatronic Units
Research on Self-ReconfigurableMechatronic Systems in MEL
Eiichi Yoshida Satoshi MurataHaruhisa Kurokawa Kohji Tomita
Shigeru Kokajihttp://www.mel.go.jp/
MELMechanicalEngineeringLaboratory
twhiteMechatronic robots: independently controlled mechatronic units. Units have memory, CPU, power etc.
Can connect, disconnect and communicate with their connected units.
• Distributed mechanical system• composed of many identical units• using local communication only• dynamically reconfigurable
• Self-assembly / Self-repair
arbitrary initial state self-assembly
self-repairusing spare unitstrouble cut off
MELMechanicalEngineeringLaboratory
Self-ReconfigurableMechatronic Systems
• Self-assembly and Self-repair
0.01.02.03.04.05.06.0
0 100 200 300 Step
differencediffusion var.
diffe
renc
e / d
iffus
ion
var.
per u
nits
remove
MELMechanicalEngineeringLaboratory
Self-ReconfigurableMechatronic Systems
Weight: 50[g] (approx. )Span: 50[mm]
SMA TorsionCoil Springs
Male:Rotating Drum
Female:Auto-locking
(releasing by SMA)
SMACoil Spring
Pin holespin
VerticalDirection
Basic Motion
MELMechanicalEngineeringLaboratory
2D Self-Reconfigurable Systems
• Experiment using 6 units
Initial Final
Moving Unit
MELMechanicalEngineeringLaboratory 2D Self-Reconfigurable Systems
twhiteVERY difficult planning problem -- exponential growth of state space with number of independent units.
MELMechanicalEngineeringLaboratory 2D Self-Reconfigurable Systems
• Basic Motion
A
MELMechanicalEngineeringLaboratory 3D Self-Reconfigurable Systems
twhiteYoshida: moves in vertical plane.2 retractable arms, connect with a lock and key mechanism.
Used to create stairs, bridges ...
MELMechanicalEngineeringLaboratory 3D Self-Reconfigurable Systems
Grey: distance from target different from zeo -> to be movedGreen: distance from target = 0 -> ok!Red: current moved unit (once at the time)
• 3D Mechanical Unit
connecting hand rotating arm span: 27.5cmweight: 6.8kgDC motor: 7W
MELMechanicalEngineeringLaboratory 3D Self-Reconfigurable Systems
twhitehttp://www-2.cs.cmu.edu/~unsal/publications/dars00_unsal_camera.pdf
Nice paper on the planning problem.
Modular Robotics (PolyPod)
Mark Yim (Stanford University, Xerox Park Research Center, 1994- …)
Parallel structure, 10 links Two types of modules: segment andnodes
• Nodes: power modules• Segments: HC11 microcontroller,
angle position (potentiometer), IR localcommunication
Modular Robotics (PolyPod)
Mark Yim (Stanford University, Xerox Park Research Center, 1994- …)
twhiteRolling ...
twhiteWalking ....
twhite
Modular Robotics (PolyBot)
Mark Yim (Stanford University, Xerox Park Research Center, 1994- …)
This generation of PolyBot included onboard computing(Power PC 555) as well as the ability to reconfigureautomatically via shape memory alloy (SMA) actuatedlatches.
The future … PolyBot
twhiteMuscle Wires and SMAsShape Memory Alloys (SMAs) are unique composites that undergo a crystaline phase change (shape change) when heated or cooled. Nitinol, a SMA combination of nickel and titanium, can be processed into various forms and is an alternative for obtaining motion
NanoMuscle™The amazing new NanoMuscle Actuator contains a long length of SMA Wire in a compact electromechanical device that amplifies the stroke and motion of the wire, and provides for easy mounting and controllability. For silent, direct linear action, grab hold of NanoMuscles.
Muscle Wires & SMAsShape Memory Alloys (SMAs) are unique composites that undergoe a crystaline phase change (shape change) when heated or cooled. Nitinol, a SMA combination of nickel and titanium, can be processed into various forms and is an alternative for obtaining motion from electrical current. Typically motors or solenoids would be used, but there are many cases when SMAs are advantageous because of their mechanical simplicity, high strength to weight ratio, silentness, and precise control.
Self-Configurable and Modular Robotics
Other researchers:• Hajime Asama, RIKEN Center, Saitama, Japan
• http://www.riken.go.jp/eng/index.html• Distributed assembling and disassembling of stair-like
structures
• Peter Will and Wei-Min Shen, ISI-USC, L.A.,US• http://www.isi.edu/conro• Modular robotics
twhiteRun video of snake-like mechatronic robot rising and others.
twhiteDescription of the CONRO system.
Crystalline Atomic Modular Self-reconfigurable Robot
• http://www.ai.mit.edu/~vona/xtal• Autonomous self-reconfiguring robots• Simulator (Xtalsim) created
– 2D and 3D• Automated planning algorithm developed
– Melt-Grow planner: O(n2) for n atoms
Conclusion
Self-Assembly, Distributed Building, Modular Robotics
Very promising domain but
• System design and hardware development are crucial
• Energetic problem still unsolved
• Scalability and distributed control still unsolvedproblems: how to control individual units forobtaining a given team performance?
Evolution and bio-inspiration can help:incremental/modular evolution, reverse engineering