research challenges in cps - artemis-ia-eu · swarm the fog . the cps ecosystem internet of things...
TRANSCRIPT
Research Challenges in CPS
Radu Grosu
Fakultät für Informatik Technische Universität Wien
1980: e.g. airbag Embedded Systems
Quick History
speed sensors
abscomputer
cruise control
distance sensors
wireless network
gps
tt-bus et-bus
1990: e.g. car airbag
1980: e.g. airbag Embedded Systems
Networked Embedded Systems 1990: e.g. car
Quick History
> 40 processors, 60 sensors, 40 actuators > 100 million lines of code controlling them
Uncertainty
Multi-scale
1980: e.g. airbag Embedded Systems
Networked Embedded Systems 1990: e.g. car
Cyber-Physical Systems 2010: e.g. smart mobility
Quick History
Time
Space
CPS Wake-Up Call
2008: NSF and US-Scientists send CPS-Manifesto to • The president of the US • President’s Council of Advisors on Science and Technology • NSF program takes off in February 2009 within the US
2012: Acatech and DE-Scientists send CPS-Manifesto to • Germany’s Federal Ministry of Education and Research • Program takes off in 2013 in Germany • H2020 program takes off in 2014 within the EU
It is high time for a Big-Push in Austria, too!
Unmanned Cars
Cyber Biological
Unmanned Trains
Where Are We Now?
Unmanned Underwater Vehicle Unmanned Aerial Vehicle
Unmanned Factory
What are the Grand Challenges? • Zero traffic-fatalities • Blackout-free electricity • Energy-aware buildings • On-the-fly production • Everywhere health-care • Max-yield agriculture
Smart City
Transportation
Energy And Water
Education
Public Safety
Social Programs
Health Care
Environment
Smart Buildings Urban Planning
Government Administration
The CPS Ecosystem
Internet
of Things (IoT)
Industry 4.0 Industrial
Internet
The Cloud Smarter Planet Machine
to Machine (M2M)
CPS
The Swarm
The Fog
The CPS Ecosystem
Internet
of Things (IoT)
Industry 4.0 Industrial
Internet
The Cloud Smarter Planet Machine
to Machine (M2M)
CPS
The Swarm
The Fog
2020:103 dev/person
The TerraSwarm
2020: 7 trillion in total
What are the Technical Challenges?
Mathematics • Discrete-continuous • Very different math
Architecture • Huge complexity • CPS-OS platform
Spacetime • Various scales of ST • ST-aware programs
Light: Particle-wave duality
Babel Tower Milan Dome
Deforming membrane
What are the Technical Challenges?
Uncertainty • Partial knowledge • Limited resources Orbitals of the hydrogen atom
Swarm Flock
1000 trillion synapses
100 billion neurons
θ0
x0
Deterministic Algorithm
while (x > x0) go back;while (θ > θ0) turn;...
Not smart/robust • Optimization tough • Not robust (slding)
Optimization goal: Learn x0, θ0, …
Capturing freedom
θ0
x0
θ0 N(θ0,σ2)
O : N(x0,σ1)
x0
Capturing freedom: Neural Network
nwhile (x > x0,σ0) go back;nwhile (θ > θ1,σ1) turn;...
Oθ(x),Oσ(x)
x0,σ1
• Ontology O: Neural Network of guard dependencies
Stochastic Algorithm
θ1 = Oθ(x)σ1 = Oσ(x)
θ0 N(θ0,σ2)
O : N(x0,σ1)
x0
Determinization Trick (e.g. In Time): Use IoI
Interval of Indifference (IoI) σ
Classification (Prediction) Limits in FOT
What about: First Order Theory (R, ≤ ,+,×,sin) ?
−Undecidable: Would allow encoding of polynomial diophantine equations (PDE)
f(x,y) > 0
f(x,y) < 0
x
y
−PDE: Known to be undecidable (Matiyasevich, 1970)
Fixing Precision (i.e. Use IoI)
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Decidable: But exponential in the number of boxes!
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
What about: FOT (R, ≤ ,+,×,sin) ?
Top-Down Algorithm
x
y
f(x,y) < 0
f(x,y) > 0
By fixing precision: Always terminates!
i Time Domain: Time triggered approach
i Value Domain: Similar to time domain abstraction
i Consequence: Uncertainty within white boxes (IoI)
What are the Technical Challenges?
Uncertainty • Partial knowledge • Limited resources
Smartness • Adapting is in big no • Neural circuits
Safety • Safety is in big no • Emergent behavior
Orbitals of the hydrogen atom
Swarm Flock
1000 trillion synapses
100 billion neurons
What About Education?
What about now? • Age of system building! • Engineering converges
It’s a Fan!
It’s a Spear!
It’s a Wall!
It’s a
Rope!
It’s a Tree!
It’s a Snake!
Back in the future (1950s) • Computation: von Neumann • Sensing/inference: Wiener • Actuation/Control: Kalman • Communication: Shannon
Recent past (1980-2010s) • Blind man assessing an elephant • Blind man building an elephant!
What is in Store for Us?
Better Health Clean Environment
Time for Family
Higher Productivity
Time for Ourselves Time for Elderly
More Vacation More Learning More Flexibility