1. Cyber Physical Systems: The Need for New Models and Design Paradigms
Bruce H. Krogh
Carnegie Mellon University

2. Cyber-Physical systems
Cyber-Physical Systems (CPS) are integrations of computation and physical processes.1
What’s new?
size and power of computational elements
pervasive networking
sensing technology
actuation technology
What’s old?
modeling and design paradigms
1 Computing Foundations and Practice for Cyber-Physical Systems: A Preliminary Report
Technical Report No. UCB/EECS , May 21, 2007
Edward Lee, University of California at Berkeley

3. More on Cyber-Physical Systems2
Some defining characteristics:
Cyber capability in every physical component
Networked at multiple and extreme scales
Complex at multiple temporal and spatial scales
Dynamically reorganizing/reconfiguring
High degrees of automation, control loops must close at all scales
Operation must be dependable, certified in some cases
Goals of a CPS research program
A new science for future engineered and monitored systems (10-20 year perspective)
Physical and cyber design that is deeply integrated
What cyber-physical systems are not:
Not desktop computing
Not traditional, post-hoc embedded/real-time systems
Not today’s sensor nets
2 CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

4. Example: Health Care and Medicine
National Health Information Network, Electronic Patient Record initiative
Medical records at any point of service
Hospital, OR, ICU, …, EMT?
Home care: monitoring and control
Pulse oximeters (oxygen saturation), blood glucose monitors, infusion pumps (insulin), accelerometers (falling, immobility), wearable networks (gait analysis), …
Operating Room of the Future (Goldman)
Closed loop monitoring and control; multiple treatment stations, plug and play devices; robotic microsurgery (remotely guided?)
System coordination challenge
Progress in bioinformatics: gene, protein expression; systems biology; disease dynamics, control mechanisms
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University
Images thanks to Dr. Julian Goldman, Dr. Fred Pearce

5. Example: Electric Power Grid
Current picture:
Equipment protection devices trip locally, reactively
Cascading failure: August (US/Canada) and October (Europe), 2003
Better future?
Real-time cooperative control of protection devices
Or -- self-healing -- (re-)aggregate islands of stable bulk power (protection, market motives)
Ubiquitous green technologies
Issue: standard operational control concerns exhibit wide-area characteristics (bulk power stability and quality, flow control, fault isolation)
Technology vectors: FACTS, PMUs
Context: market (timing?) behavior, power routing transactions, regulation
IT Layer
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University
Images thanks to William H. Sanders, Bruce Krogh, and Marija Ilic

6. Pervasive Underlying Problems, Not Solved by Current Technologies
How to build predictable real-time, networked systems at all scales with integrated models of the physical world?
How to formulate and manage high-confidence, dynamically- configured CPS?
How to organize inter-operable “aggregated” systems?
How to cooperatively detect and manage interference among systems in real time, avoid cascading failure?
How to formulate an evidential (synthetic and analytic) basis for trusting systems?
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

7. Impending Technical Challenges
Shift FROM
compartmentalized designs of physical systems, control subsystems and software architecture
lack of knowledge on the cyber side of engineering principles and physical laws (and vice-versa)
cyclic executives + human- and information-centric operation
centralized
separation in time and space TO
integrated and optimized design
CPS-awareness and expertise
to highly-automated, autonomous, coordinated frameworks
to federated, decentralized, open and configurable
multi-scale systems, mixed synchronous/reactive systems
Still
real-time (perhaps wide-area, time-critical), still safety- and security-critical, require certification
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

8. Recent Workshops on Cyber-Physical Systems
“High Confidence Medical Device Software and Systems (HCMDSS)”, June 2 - 3, 2005, Philadelphia, PA
“Aviation Software Systems: Design for Certifiably Dependable Systems”, October 5-6, 2006, Alexandria
NSF Workshop on “Cyber-Physical Systems”, October 16-17, 2006, Austin,
“Beyond SCADA: Networked Embedded Control for Cyber Physical Systems (NEC4CPS)”, November 8 & 9, 2006, Pittsburgh
“High-Confidence Software Platforms for Cyber-Physical Systems (HCSP-CPS), November 30 – December 1, 2006, Alexandria
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

9. Industry Round-Table on CPS NSF, May 17, 2007
Health-Care
Doug Busch, VP and CTO of Digital Health Group, Intel
David R. Jones, Director Quality Assurance, Regulatory Affairs and Philips Business Excellence, Philips Consumer Healthcare Solutions
Automotive Systems
Nady Boules, Director, Electrical and Controls Integration, General Motors
Venkatesh Prasad, Director, Ford
Building and Process Controls
J. Michael McQuade, Senior VP, Science and Technology, United Technologies
Steve Schilling, VP, Emerson Process Control
Defense and Aviation Systems
John Borgese , VP of Advanced Technology Center, Rockwell Collins
Gary Hafen, Director of Software Engineering, Lockheed Martin Corporate Headquarters
Peter Tufano, VP of Engineering for Network Enabled Systems, BAE
Don Winter, VP of Engineering and Information Technology, Boeing PhantomWorks
Critical Infrastructure
Guido Bartels, Director, IBM Global Energy and Utility Solutions
Henry Kluepfel, Vice-President, SAIC
Venture Capital
David Tennenhouse, General Partner, New Venture Partners
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

10. Design of Embedded Control Systems
Traditional approach: Separation of Concerns
Control-theoretic design of continuous dynamic feedback loops
ignore implementation details: mode switching, fault detection, real- time constraints, implementation platform, etc.
Event-based design to supervise real-time control loops
ignore continuous dynamics: stability, transient response, parametric variations, etc.

11. Design of Embedded Control Systems
Traditional approach: Separation of Concerns
Control-theoretic design of continuous dynamic feedback loops
ignore implementation details: mode switching, fault detection, real- time constraints, implementation platform, etc.
Event-based design to supervise real-time control loops
ignore continuous dynamics: stability, transient response, parametric variations, etc.
This works in most cases, BUT ...

12. Demands from Emerging Applications
New challenges
increasingly complex applications
safety critical systems
autonomy
multi-agent
increasingly complex solutions
heterogeneous, distributed platforms
sophisticated numerical control algorithms
Implications
engineering insight is inadequate
testing-based V&V is insufficient
move toward model-based design

13. Tools for Design & Implementation of Embedded Control Systems
Control Implementation: Discrete State/Events
Lyapunov functions, eigenspace analysis, etc.
Analytical Tools
MATLAB, MatrixX, VisSim, etc.,
Software Tools
Control Design: Continuous State
differential equations, transfer functions, etc.
Models
automata, Petri nets, statecharts, etc.
Boolean algebra, formal logics, recursion, etc.
SCADE, Statemate, SMV, SAT, etc.

14. Limitations of Conventional Control System Design (CCSD)
Inputs/outputs are not intrinsic
From following commands to implementing intent
Human-system interaction
Deeply embedded CPS

15. Inputs/outputs are not intrinsic
CCSD assumes an I/O structure. In CPS, the identity of input/output signals is context dependent (at best).
steer-by-wire
temperature
door closer
(J. C. Willems)

16. Inputs/outputs are not intrinsic
CCSD assumes an I/O structure. In CPS, the identity of input/output signals is context dependent (at best).
steer-by-wire
Model context-dependence
as hybrid systems w/
mode switching
temperature
door closer
(J. C. Willems)

17. Inputs/outputs are not intrinsic
CCSD assumes an I/O structure. In CPS, the identity of input/output signals is context dependent (at best).
steer-by-wire
Physical modeling “languages”:
bond graphs
Omola/Dymola
SimMechanics
temperature
door closer
(J. C. Willems)

18. From following commands to realizing intent
CCSD assumes command-following performance measures. CPS will realize the intent of the user.
ABS
Automated External Defibrillator
power grid?

19. From following commands to realizing intent
CCSD assumes command-following performance measures. CPS will realize the intent of the user.
ABS
Integration of logic/rules/events with continuous/timed feedback control
(hybrid systems)
Automated External Defibrillator
power grid?

20. From following commands to realizing intent
CCSD assumes command-following performance measures. CPS will realize the intent of the user.
ABS
Automate system operation under stressed conditions.
Automated External Defibrillator
power grid?

21. Human-system interaction
CCSD assumes only information feedback.
CPS will include physical feedback.
building control?
aircraft
ABS
Boeing 777
Airbus 380

22. Human-system interaction
CCSD assumes only information feedback.
CPS will include physical feedback.
Haptic systems design
building control?
aircraft
ABS
Boeing 777
Airbus 380

23. Human-system interaction
CCSD assumes only information feedback.
CPS will include physical feedback.
building control?
aircraft
ABS
Integrate human behavior into the control loop (e.g., make it uncomfortable so they will open the windows)
Boeing 777
Airbus 380

24. Deeply embedded CPS
In CCSD embedded components close local “inner” feedback loops.
CPS will enhance and leverage nature physical feedback at all levels.

25. E.g., medical implants that work with the natural healing processes
Deeply embedded CPS
In CCSD embedded components close local “inner” feedback loops.
CPS will enhance and leverage nature physical feedback at all levels.
E.g., medical implants that work with the natural healing processes

26. Physical is central to CPS:
We need
new cross-cutting paradigms
new architectures
CPS will lead to
more rapid transition of science/technology to critical applications

27. Possible Grand Challenges3
Zero automotive traffic fatalities, injuries minimized, and significantly reduced traffic congestion and delays
Blackout-free electricity generation and distribution
Reduce testing and integration time and costs of complex CPS systems (e.g. avionics) by one to two orders of magnitude
Perpetual life assistants for busy, older or disabled people
Extreme-yield agriculture
Energy-aware buildings
Location-independent access to world-class medicine
Physical critical infrastructure that calls for preventive maintenance
Self-correcting and self-certifying cyber-physical systems for “one-off” applications
3 Industry Roundtable on Cyber-Physical Systems
NSF, May 17, 2007
Raj Rajkumar, Carnegie Mellon University

28. Cyber Physical Systems or Cyber for Physical Systems
How should the requirements for control (and other) physical applications influence “cyber” research?
Will the standard separation of concerns approach (applications vs. computing infrastructure) continue to work well?

29. Issues in Education computer science domain experts (engineers)
focuses on discrete mathematics
little emphasis on numerical methods
limits the understanding of physical systems
domain experts (engineers)
focuses on mathematics for analysis and design
little exposure to embed and real-time computing
limits the understanding of real-time implementation
We need to re-think how we educate domain experts and computer scientists if we are going to realize
sustainable CPS.

30. Core CPS Programmatic Themes
Scientific foundations for building verifiably correct and safe cyber-physical systems
Scalable infrastructure and components with which cyber-physical systems can be deployed
Tools and Experimental Testbed
Education that encompasses both the cyber and the physical domains
CPS Briefing
NSF, May 10, 2007
Raj Rajkumar, Carnegie Mellon University

31. Long-Term CPS Goal
Transform how we interact with the physical world just like the internet transformed how we interact with one another.
Convergence of embedded systems, control theory, hybrid systems, microcontrollers, sensors, actuators, wireless networks, wide area networks, distributed systems, operating systems, advances in structures, …
Seek scientific foundations and technologies to integrate cyber-concepts with the dynamics of physical and engineered systems.
Industry Roundtable on Cyber-Physical Systems
NSF, May 17, 2007
Raj Rajkumar, Carnegie Mellon University