Additional Notes: 1.Retrofitting Legacy systems 2.Market lifecycle issues that are embedded in the systems 3.Need better measurements and metrics 4.Intractable where to start accretion process, high gain vs. low gain 5.HUMAN side. Humans are unpredictable but there may be some consistencies. Can we do, for example, data analyses through heat mapping? And how do you display info so that it is easily perceived? Need actionable intelligence. Must be able to extrapolate though modelling 6.Need modelling and simulation tools to design strategy: Operational aspects (not just normal and emergency, but have various levels of emergency, emergency level 1 and level 2, for design and planning. Based on variable uncertainty of knowledge and number of sensors. This should affect our decisions with respect to sensor mix and optimal placement.
Additional Notes: 7. Speed & feed of networks; Robustness; Security; Privacy (may be less important to future generations. e.g.., Facebook) 8. Need TOOLS for abstracting and expressing emergent properties 9. Need architectural principles for embedded networks. We can make local networks application oriented. We could decouple hardware & software for a local network. This should reduce cost and increase productivity. 10. How can we ascertain that chips and routers can be trusted? Chips could be hacked before they leave the factory. 11. How do we do better transition and how do we get a fundamental research agenda? These are different questions for cyber and for physical systems because we have actuation. Don’t know how to do V&V (verification & validation). The big problems have both continuous and discrete aspects and there are no V&V tools.
Additional Notes: 12. What are the consequences of feedback and actuation in a network setting? What are the consequences of the design tradeoffs that we make? What qualities make things better and what qualities make it worse? Do we know this? Can we measure? 13. How does risk and uncertainty propagate? Why does uncertainty sometimes propagate a lot and sometimes very little? 14. This may be connected to bio systems which have low gain and high gain systems. Very large disturbance can make only a little difference. The brain regulates. But if I build a high gain system it could be hijacked and the immune system goes wrong and you die. 15. Need robust performance in view of possible failure modes. The smart grid could be hacked and possible create artificial demand. Then it could be turned off. Thus, could create system oscillations. Topology can amplify or dampen oscillation.
Additional Notes: 16. Our aircraft are 70-80% software in cost & testing and recurring cost. This is important from a resource and vulnerability standpoint. More & more you can’t tell what the infrastructure will do next. Stochastic properties are where the next conceptualization is headed. There was a conference in Europe using applied category theory to predict complex system behavior. 17. Building gateways and sensors: there is no commonality between communications capabilities and sensor capabilities. Need homogeneity among sensor protocols. Provisioning – need common way for sensors to join networks and then deregulate them when they are finished. (We need a definition of these terms in the report.) 1) Need work on time mechanisms (actuators?) 2) Need a formal management system of resources across network) 3) Need formal methods for non-invasive instrumentation.
18. Need architecture design principles for safe and secure systems. E.g.. neocortex says breathe I and breathe out. Can we put these principles into network design so as to stay within high and low coordinates, and thus make the networks tamper and attack resistant. 19. We still lack good models. Can we make more advanced tools to measure uncertainty? How about an approach based on measure theory? Want to mix atomic and non-atomic scale distributions coexisting in the same objects. 20. Also need to learn how to use these tools on REAL networks. Need reasonable computational tools. 21. Apply mathematics that has not been traditionally thought of as applied. Mix reals and Booleans. 22. Want a science of complex engineered systems and cybersecurity. Need to understand the mathematics of complex cyber-physical systems and design for optimal robustness and resilience. Additional Notes:
Test beds: Smart grid and intelligent traffic control Need a UNIFIED THEORETICAL FOUNDATION for different network structures, e.g.. interdependence of different components of the power grid and cybersecurity. Should show the interdependence across different subsystems and how to capture this info. And social network behavior. What is the right abstraction? We need the right language of architecture as constraints and standards. Need some underlying principles. As we build bigger systems, the human ability to understand is limited. Need to be able to make decisions at the speed of a machine. Need to do it automatically. Need self-optimizing and self-healing. Smart Grids, Power and Energy: We are losing $18-20 billion per year by not optimizing these systems. In addition, a minimum 4-5% of efficiency increase ($20.4 billion annually) is within reach. $80-188 billion lost in annual outages -- can be reduced by $49 billion. Get an optimal AC topology. The same optimization model can be used. There has been a resurgence to do this now. The markets would function better if we did this. Issue: Hard to do globally.
Cyber-Physical Networks -- Summary 1.Hard to distinguish different areas of application, and hence the need for definition and characterization of Cyber-Physical Networks regimes. 2.Need for metrics and mathematical foundations for understanding, synthesizing and evaluating architectures in a systematic way, including issues such as robustness, resilience, security, reliability, fragility, and volatility, all of which need formalisms and validation. 3.Tools and methods, mathematical and constructive frameworks, for understanding, designing and conducting tradeoffs between the various characteristics of the system, including suitability for use, economics considerations and ROI, performance, etc.
Cyber-Physical Networks – Summary (cont.) 4. When it comes to specific application, creating systems and methods for non-intrusive instrumentation and testing – Improvement of mathematical methods to provide robust tools for integrated cyber-physical V&V – Use examples from other pertinent fields, e.g.. biologically inspired architectures and applications – Learning algorithms for improved performance in actual use 5. There are examples of what our nation is significantly investing in; these are opportunities for focus: – Smart Grid (incl. integrated models, architectures, sensors, measurement, methods, mgmt.) – Sensor networks with diverse applications (incl. defense, industrial instrumentation, etc.) – Broad applications such as telematics, healthcare, emergency communications – Specific test beds such as Genie
Question 1: Definitions *** ENGINEERED Engineered network: one that delivers services in a dependable ways: e.g. –Bits (e.g. Internet) –Information (e.g. Bloomberg) –Knowledge (on top of social networks) –Decision networks (e.g. manage a complex workflow, logistics). *** * PHYSICS What distinguishes CPN they have to provide services to components of the systems constrained by the laws of physics. It emphasizes the way the networks has to self-function. **** CYBER : considerable requirement for storage, computing, and communications, and actuation of the underlying substrate ( electronics, mechanical, etc.)
Question 1: Dimensions *** DECISION PARADIGM Range of system from 1 decision maker very tightly integrated to distributed decision traffic control is a network. **** RESPONSE TIME : distinguish between time critical and opposite. E.g. a CAR, SCADAS, public networks have different requirement. **** USE System built for limited loads (e.g.. airplane, stopped when outside the environments); and ones very actuated with hard to predict loads and perturbations. **** SPATIAL EXTENT : all the way from global down to smallest level. Research collaboration networks (control an experiment), Climate data networks,
Wishlist & Moves – Question 2 Short Term Moves – 1-3 years whatWho / HowOutcomes Recast known systems as proto-architectures: Turing, Shannon, Bode, Chris Alexander, Bio NSF, DARPA, ONRUnique framework and language For heterogeneous architectures/models/components Medium Term Moves – 4-5 years WhatWho / HowOutcomes Flexible architecturesDOE, AFOSR, DHSMetrics, theoretical bounds, data-model positive spiral Long Term Moves – 6-10 years WhatWho / HowOutcomes New paradigm for CES: New math, hybrid systems EveryoneOptimize CES functioning
Wishlist & Moves– Questions 3-4 Short Term Moves – 1-3 years whatWho / HowOutcomes Management of expectations of R&D process Shared between the R&D community and sponsorsCategorization and definition of goals for different modeling activities Case studies of complex engineering networks Interdisciplinary and multi-component, i.e., academia, industry, and sponsorship Fostering partnerships and defining guidelines Generating intuition and building knowledge base and datasets Generate interdisciplinary collaborations sponsorsForming communities Medium Term Moves – 4-5 years WhatWho / HowOutcomes building tools for modeling, validation and simulation Shared between the R&D community and sponsorsDeep insight and understanding of the system, acceleration of system design and analysis, techniques for model abstraction Modeling techniques to capture the interdependency across subsystems Shared between the R&D community, industry and sponsors
Wishlist & Moves– Questions 3-4 Long Term Moves – 6-10 years WhatWho / HowOutcomes Open source Models for risk and uncertainty Understand the propagation of risk and uncertainty across subsystems and scales Common modelling framework Optimum Operation Strategy for CPS
Wishlist & Moves -- Applications Short Term Moves – 1-3 years whatWho / HowOutcomes Testbed for the last mileDesign the last mile Testbed for Self-healing Smart Grid Private/Public: DOE/FERC/DHS and private sector, Univ.: Distribution Automation Systems, AMI with full cyber security Architecture to reduce outages and PQ disruptions by XX% Cyber-physical secure with full customer privacy Testbed for HealthcareGoal: Robust diagnostics from the start (e.g.., case study Henry Ford Hospital) Cyber secure patient-centered care delivery system See “Watson" Medium Term Moves – 4-5 years WhatWho / HowOutcomes Testbed for Self-healing Smart Grid Reduce costs by xx%-- do it efficiently and optimally Testbed for HealthcareRobust diagnostics from the start (e.g.., case study Henry Ford Hospital) Cyber secure patient-centered care delivery system See “Watson” Long Term Moves – 6-10 years WhatWho / HowOutcomes Testbed or the last mileX,XXX two-way video channels to your home Testbed for Self-healing Smart Grid A self-optimizing, self-reconfiguring cyber-physical system