Engineering Complexity

8
Engineering Complexity Engineering Complexity Nicolas Demassieux 1 © Nicolas Demassieux

description

A high level presentation about the way traditional engineering methods need to change, in order to manage the design and operation of increasingly complex technical systems. A parallel with "engineering methods" of living ecosystems is used.

Transcript of Engineering Complexity

Page 1: Engineering Complexity

Engineering ComplexityEngineering Complexity

Nicolas Demassieux

1© Nicolas Demassieux

Page 2: Engineering Complexity

Complexity of Interacting Complexity of Interacting SystemsSystems

Internet SecurityInternet Security Hubble mirrorHubble mirror ARIANE 5 first flight ARIANE 5 first flight –– Space shuttleSpace shuttle

2© Nicolas Demassieux

Page 3: Engineering Complexity

Engineering Engineering EnvironmentEnvironment

PASTPAST Simple predictable Simple predictable environmentenvironment

Low interactionsLow interactions Low complexityLow complexity Expensive ControlExpensive Control Cheap EnergyCheap Energy A priori validationA priori validation Strict plans (“design”)Strict plans (“design”)

FUTUREFUTURE Complex environmentsComplex environments Complex interactionsComplex interactions Really High complexityReally High complexity Cheap ControlCheap Control Costly EnergyCostly Energy In situ validationIn situ validation Indirect plans Indirect plans (“specification”)(“specification”)

© Nicolas Demassieux

Page 4: Engineering Complexity

Life “know-Life “know-how”how”

CELLCELL Self Self AssemblyAssembly

Self RepairSelf Repair Self Self RecyclingRecycling

DNADNA RobustnesRobustnesss

ReplicatiReplicationon

EvolutionEvolution

ORGANISMORGANISM Energy Energy SupplySupply

IntelligenceIntelligence AdaptationAdaptation

ECOSYSTEMSECOSYSTEMS Co-evolutionCo-evolution DiversityDiversity Long term Long term survivalsurvival

4© Nicolas Demassieux

Page 5: Engineering Complexity

EngineerinEngineeringg

SUBSYSTEMSSUBSYSTEMS Self Assembly : Self Assembly : Boot, Pilot Boot, Pilot

channelschannels, , Molecular Molecular engineering,engineering, Quantum dots, Quantum dots, Nanotubes, Shape memory Nanotubes, Shape memory alloys, ...alloys, ...

Self Repair : Self Repair : Disk Disk unfragmentation,unfragmentation, Self check, Self check, Fault tolerance, ...Fault tolerance, ...

Self Recycling : Self Recycling : Garbage Garbage collectors,collectors, Self disassembly, Self disassembly, biodegradable materials ...biodegradable materials ...

DESIGNDESIGN Robustness : Robustness : DIGITALDIGITAL Replication : Replication : SoftwareSoftware Objects Objects

(Inheritance)...(Inheritance)... Evolution : Evolution : Monte Carlo,Monte Carlo, Genetic Algorithms, Genetic Algorithms,

Artificial life...Artificial life...

SYSTEMSSYSTEMS Energy Supply :Energy Supply : Intelligence : Intelligence : Neural

networks, Agent technology, Cellular Automatas, AI...

Adaptation : Adaptation : Self Self optimizing systems, optimizing systems, Software radios, Software radios, Adaptive techniquesAdaptive techniques

TECHNOSPHERETECHNOSPHERE Co-evolution : Co-evolution : the the

Internet, Radio Internet, Radio ecosystems, service ecosystems, service pyramid….pyramid….

Diversity : Diversity : Non Non convergenceconvergence

Long term survival : Long term survival : Non Non optimalityoptimality

5© Nicolas Demassieux

Page 6: Engineering Complexity

One Example : OptimisationOne Example : Optimisation

CONTINUOUS CONTINUOUS OPTIMISATIONOPTIMISATION

Energy landscapeEnergy landscape Local minimaLocal minima Gradient Gradient

algorithms, algorithms, simulated simulated annealingannealing

BOOLEAN OPTIMISATIONBOOLEAN OPTIMISATION

Boolean function : {0,1}Boolean function : {0,1}n n

{0,1}{0,1} Problem : Synthesis using Problem : Synthesis using

boolean gates with minimal boolean gates with minimal cost …. a NP hard problemcost …. a NP hard problem

Solutions : Iterative Solutions : Iterative heuristics, genetic heuristics, genetic programming, ...programming, ...

0 1

01

1 0

10

ABC 6© Nicolas Demassieux

Page 7: Engineering Complexity

One Example : Optimisation (2)One Example : Optimisation (2)

COMPLEX MULTIPARAMETER SYSTEMS COMPLEX MULTIPARAMETER SYSTEMS CONTROLCONTROL

Problem : “tweak the knobs of Problem : “tweak the knobs of a system” to optimize its a system” to optimize its behavior in a given complex behavior in a given complex environmentenvironment

Solutions : genetic Solutions : genetic programming, optimal rate programming, optimal rate distorsion, self optimizing distorsion, self optimizing systemes, self awareness...systemes, self awareness...

DCT Q

image

reconstruite

image

source VLC

mémoired'image

+

-

++ Q

-1DCT

-1VLC

-1

mémoire

d'image

estimationmouvement

Q-1

DCT-1

Buffer

Buffer

COMPLEX MULTIDIMENSIONAL COMPLEX MULTIDIMENSIONAL “SHAPES” IDENTIFICATION“SHAPES” IDENTIFICATION

Problem : Classification, Problem : Classification, learninglearning

Solutions : VQ, Neural Solutions : VQ, Neural network, fractal ...network, fractal ...

7© Nicolas Demassieux

Page 8: Engineering Complexity

New Engineering New Engineering CultureCulture

PASTPAST Models, black BoxesModels, black Boxes ““Hard Specifications”Hard Specifications” OptimizationOptimization Engineered controlEngineered control Independent SystemsIndependent Systems ConfidenceConfidence

FUTUREFUTURE No more complete modelsNo more complete models Evolving specificationsEvolving specifications OverprovisionOverprovision Autonomous controlAutonomous control Interdependent SystemsInterdependent Systems HumilityHumility

Engineering Optimized Engineering Optimized SystemsSystems

Managing Technical Managing Technical

EnvironmentsEnvironments8© Nicolas Demassieux