480 likes | 748 Views
Complex Systems Phenomena, Characteristics & Research Questions. William B. Rouse. Overview. Complexity & Complex Systems Definitions Characteristics Views Studies of Complexity Fault Diagnosis Large-Scale Systems Disease Control Workshop on Complex Systems Motivation Agenda Results
E N D
Complex SystemsPhenomena, Characteristics & Research Questions William B. Rouse
Overview • Complexity & Complex Systems • Definitions • Characteristics • Views • Studies of Complexity • Fault Diagnosis • Large-Scale Systems • Disease Control • Workshop on Complex Systems • Motivation • Agenda • Results • Recommendations • Agent-Based Modeling • Opportunities • Challenges • Needs • Summary
Complexity • The intrinsic amount of resources, for instance, memory, time, messages, etc., need to solve a problem or execute an algorithm (NIST, 2004) • How long it would take, or how much capacity would be required, at a minimum, for a standard universal computer to perform a particular task (Gell-Mann, 1995) • The length of a concise description of a set of an entity’s regularities (Gell-Mann, 1995) • The essence of complexity is the elaboration of highly structured communication, computing, and control networks that also create barriers to cascading failure events (Carlson & Doyle, 1995)
Complex Systems • A system with a large number on mutually interacting dynamical parts which are coupled in a nonlinear fashion (Wikipedia, 2006) • Relationships are nonlinear • Relationships contain feedback loops • Complex systems are open • Complex systems have a memory • Complex systems may be nested • Complex systems have networks of dynamic relationships • Complex systems can include humans and organizations • Boundaries of complex systems are difficult to determine
Complex Systems • System: A group or combination of interrelated, interdependent, or interacting elements that form a collective entity. Elements may include physical, behavioral, or symbolic entities. Elements may interact physically, mathematically, and/or by exchange of information. Systems tend to have purposes, although in some cases the observer ascribes such purposes. • Complex System: A system whose perceived complicated behaviors can be attributed to one or more of the following characteristics: large numbers of elements, large numbers of relationships among elements, nonlinear and discontinuous relationships, and uncertain characteristics of elements and relationships. Complexity is perceived because apparent complexity can decrease with learning.
Characteristics • Number of Elements • Number of Relationships • Nature of Relationships • Logical: AND vs. OR & NAND • Functional: Linear vs. Nonlinear • Spatial: Lumped vs. Distributed • Structural: Feedforward vs. Feedback • Response: Static vs. Dynamic • Time Constant: (Not Too) Fast vs. (Very) Slow • Uncertainty: Know Properties vs. Unknown Properties • Knowledge, Experience & Skills • Relative to All of the Above • Relative to Observer’s Intentions
System (S) Complexity = f (Intentions) Input (U) Output (Y)
Views of Complex Systems • Hierarchical Mappings • Uncertain State Equations • Discontinuous, Nonlinear Mechanisms • Autonomous Agents
Hierarchical Mappings • Systems engineering = Processes for designing, developing, deploying, and sustaining complex systems • Hierarchical decomposition of a very complicated design task into component tasks • Management of the execution of these tasks and integration of task outcomes • Complexity typically due to large numbers of interacting elements • A large number of reasonably straightforward tasks whose outcomes will flow together to create a successful complex system • Appropriate resolution of multi-attribute tradeoffs across multiple stakeholders • Complexity managed by dividing and conquering it
State Equations • Systems engineering = Design of mechanisms whereby the “state” evolves affecting system response and stability • Central design issue is nature of appropriate feedback mechanisms for controlling system state • Observability and controllability are key constructs; optimization of control often an overriding goal • Inabilities to fully specify state-transition mechanisms & uncertainties limit formulation to constrained optimality. • Formal depiction and manipulation of mechanisms underlying complex behaviors seldom “scale up” • Complexity due to large numbers of state variables and significant levels of uncertainty • Pursuit of optimal control solutions often made possible by assumptions of linearity
Nonlinear Mechanisms • Simple underlying phenomena yield complex behaviors for systems with very few elements, perhaps even just one element with particular interaction terms • Nonlinear and/or discontinuous nature of the elements lead to behaviors labeled as catastrophes, chaos, etc. • Systems that appear simple can produce very complex behaviors; complex phenomena may be attributable to simple mechanisms. • Complexity due to departures from our expectations of continuous, linear phenomena • Understand complexity by exploring underlying mechanisms which may lead to design solutions. • Formal systems approaches tend to flounder when addressing fairly small numbers of nonlinear mechanisms
Autonomous Agents • Composition of large numbers of simple behaviors into overall system behaviors that exhibit hallmark characteristics of complex systems • Simple behaviors created by autonomous “agents” acting independently in pursuit of their individual goals • Reactions of agents to each other’s behaviors result in emergent phenomena that could not have been predicted by dissecting individual agents. • Understanding the nature of incentives, motivations, and prohibitions that will influence individual agents to contribute to creating desirable collective behaviors • Understanding and managing complexity are experimental rather than axiomatic undertakings • Many things can be demonstrated but few can be proven
An Example • Effects of turbulent flow on aerodynamic behavior and vehicle performance in high-density traffic • View No. 1 for designing the vehicle • View No. 2 to explore vehicle dynamics • View No. 3 to model turbulence • View No. 4 to understand traffic effects • Problem, e.g., poor vehicle handling qualities vs. traffic congestion problems
Studies of Complexity • Fault Diagnosis • Large-Scale Systems • Disease Control • Health Advisor
Fault Diagnosis • Finding Faults in Complex Systems • Time until correct diagnosis • AND and OR relationships • Feedback loops • Measures of Complexity • Number of components • Optimal solution • Number of relevant relationships • Information Theory metric • Evaluation (N=88 aircraft mechanics) • No. of components (r = 0.25) • Optimal solution (r = 0.50) • No. of relevant relationships (r = 0.80) • Information theory metric (r = 0.84)
Large-Scale Systems • Monitoring & Control of Large-Scale Networked Systems • Routing of requests for communications network resources • Maximize number of customers served & minimize average service time • Node failures can lead to cascading network failures • Measure of Complexity • Structural – Relates to Physical Characteristics of Network • Strategic – Relates to How Humans Address Network Performance • Evaluation • Complexity increases time & percent correct for failure diagnosis • Complexity increases with number of levels in network • Complexity increases with extent of network redundancy • Conclusions • Complexity due to dynamic interaction of structure & strategy • Aiding needed when automation masks failures
Disease Control • Stakeholders & Issues • Disease Detection • Complex Adaptive Systems • Health Advisor
Public Awareness Public Readiness Screening Available Costs Covered Public Communication Public Education Physician Education Consumer Advocacy Medical Research Disease Detection Screening Effective $ $ $ $ $ Public, Delivery System, Government, Non-Profits, Academia, Business
Complex Adaptive Systems • They are nonlinear, dynamic and do not inherently reach fixed equilibrium points. The resulting system behaviors may appear to be random or chaotic. • They are composed of independent agents whose behavior can be described as based on physical, psychological, or social rules, rather than being completely dictated by the dynamics of the system. • Agents' needs or desires, reflected in their rules, are not homogeneous and, therefore, their goals and behaviors are likely to conflict -- these conflicts or competitions tend to lead agents to adapt to each other's behaviors. • Agents are intelligent, learn as they experiment and gain experience, and change behaviors accordingly. Thus, overall systems behavior inherently changes over time. • Adaptation and learning tends to result in self-organizing and patterns of behavior that emerge rather than being designed into the system. The nature of such emergent behaviors may range from valuable innovations to unfortunate accidents. • There is no single point(s) of control – systems behaviors are often unpredictable and uncontrollable, and no one is "in charge." Consequently, the behaviors of complex adaptive systems usually can be influenced more than they can be controlled.
WORLD • Economy, e.g., Recession Decreased Sales • Politics, e.g., War Increased Acute Health • Environment, e.g., Pollution Increased Chronic Health • CONSUMER No. 1 • Health State Provider Choice • Benefits Provider Choice • Information Provider Choice • HEALTH • Probability of Chronic Problems • Probability of Acute Problems • PROVIDER No. 1 • Patients Outcomes • Patients Claims • Claims Revenue • INFORMATION • Provider Prices • Provider Performance • Symptoms Diagnosis • Diagnosis Treatment • Patient Health Record • Wellness Programs Health • CONSUMER No. 2 • Health State Provider Choice • Benefits Provider Choice • Information Provider Choice • PROVIDER No. 2 • Patients Outcomes • Patients Claims • Claims Revenue • CONSUMER No. 3 • Health State Provider Choice • Benefits Provider Choice • Information Provider Choice • PROVIDER No. 3 • Patients Outcomes • Patients Claims • Claims Revenue • BENEFITS • Family Coverage, e.g., Employee % • Medical Coverage, e.g., 80/20% • Out-of-Pocket, e.g., Co-Pay • Wellness Coverage • CONSUMER No. M • Health State Provider Choice • Benefits Provider Choice • Information Provider Choice • PROVIDER No. N • Patients Outcomes • Patients Claims • Claims Revenue • CHOICES • Coverage Yes/No • Provider 1, 2, 3 or N • INSURER • Benefits Revenue • Claims Costs • Profits Rates • EMPLOYER • Sales Jobs • Sales Benefits • Rates Benefits
Complex Systems Workshop • Motivation • Workshop Agenda • Workshop Results • Recommendations
Workshop Motivation • National Challenges, e.g., • Infrastructure • Healthcare • Environment • Security • Increased Emphasis • Systems Engineering • Engineering Systems • Systems Biology • “The World Is Flat” • 1,000,000 vs. 65,000 Engineers/Year • American Competitiveness Initiative • Intended to Double NSF Budget
NSF Workshop Agenda • September 28th • NSF Introduction – Charge to Participants (Mario Rotea) • NSF Overview – “Trends, Organization & Themes” (Richard Buckius) • Plenary Overview – “Complexity & Complex Systems” (Bill Rouse) • Context Problems Subgroups – 1st Meeting • Plenary Interim Reports on Issues Underlying Complexity • Context Problems Subgroups – 2nd Meeting • September 29th • Context Problems Subgroups – 3rd Meeting • Plenary Final Reports on Research Topics • Plenary Overview – “Complexity” (John Doyle) • Plenary Drafting of Overall Outcomes • Discussion of Next steps
Context Problems • Infrastructure & Transportation • Health Care Delivery • Bacteria Level Design • Others Considered • Environmental Control • Real Time Electronic Medical Records • Vehicle Design
Workshop Results • The complexity of a system (or model of a system) is related to: • The intentions with which one addresses the system (real vs. perceived?) • The characteristics of the representation that appropriately accounts for the system’s boundaries, architecture, interconnections, and information flows • There are usually multiple representations of a system, all of which are simplifications; hence, complexity is inevitably underestimated • The context, multiple stakeholders and objectives associated with the system’s development, deployment, and operation • The learning and adaptation exhibited during the system’s evolution • Fundamental complex systems research should focus on: • The full nature of design objectives for such systems (as known in time) • Approaches to modeling systems relative to these objectives • Methods and tools for model development and use • Means for evaluation and experimentation with models and real systems • Approaches to decision support for those who invest in, develop, operate, and use complex systems
Recommendations • Phenomena of Particular Interest • Important Characteristics of Systems • Important Research Questions
Phenomena of Interest • Human & Social Behaviors in Complex Systems • Human performance, mental models, social networks, etc. • Complex Physical Systems • Biology, ecology, weather, etc. • Interdependencies Across Scales & Domains • Time & spatial scales • Rapid Change & Uncertainties • Endogenous environment, e.g., technology • Exogenous environment, e.g., economy • Match of demands and system performance & capacity • Boundaries of System • Relationships to authority to allocate resources , determine incentives, and control in general
Team Model Trainer • Aegis Cruiser Shot Down an Iranian Airliner • Diagnosis • 25 Guys Face Armageddon • Two Times at Bat/Day – “Don’t mess up.” • Solution • More Time at Bat (50 vs. 2) • Focus on Team Mental Models • Agent-Based Simulation of Teammates
Two Theories • Sources of Denominationalism (Neibuhr, 1929) • Castes make outcastes • Outcastes make castes • Attribution of Attitudes (Jones, 1967) • Fundamental attribution error • Model-Based Dynamics • Cooperation • Conflict • Cooperation
Outcaste N+1 Caste 1 Caste 2 Outcaste N+M Caste N Caste N + 1 Outcaste N +2
Important Characteristics • Emergent Behaviors & Unintended Consequences • Robustness, Resilience, Flexibility, Agility, Adaptability, Evolvability, Etc. • Fundamental Limits • Information access • Knowledge of system • Well posedness of models • Design practices • Nature of “state” • Observability of states • Controllability of states • Scalability
Research Questions • What architectures underlie (physical, behavioral & social) phenomena of interest? (Scientific explanation) • Conceptual frameworks, representations, structures, models, etc. • How are architectures a means to desired system characteristics? (Engineering methodology) • Modeling vs. sensing; harmonization; economics of architectures • How can architectures enable resilient, adaptive, agile, evolvable systems? What is fixed and what changes? • How can you (empirically) evaluate & assess architectures prior to and subsequent to development & deployment? • What is the nature of fundamental limits of information, knowledge, model formulation, observability, controllability, scalability, etc.?
Definitions • Architecture – Include Illustrations • IEEE Std 1471, DoDAF, SoA, etc. • Information systems, vehicles, enterprises, etc. • Product vs. process • Complexity, Complex Systems, System of Systems, Etc. • Elaborate of use of the term architecture vs. architecture frameworks, conceptual frameworks, representations, models, etc. • Emphasize entities, relationships, behaviors, and performance.
R&D World $ R&D characteristics • Value streams • Multi-stage decision process • Uncertainty, market risk, technical risk • Deployment, deferral, obsolescence R&D World analysis • How should R&D be valued (e.g., real options vs. NPV)? • How should budget be allocated over R&D stages? • What is the effect of uncertainty, market risk and technical risk and delays? • What is the accuracy of analytic models vs. simulation? Extensions for social networks (IBM) $ Stage 1 $ Stage 2 Not funded $ Stage 3 Technical failure Stage 4 Value realized
Agent-Based Modeling • Opportunities • Challenges • Needs • Data Sets • Methods & Tools • Testbeds
Opportunities for ABM • Healthcare, Infrastructure, Security, Etc. Critical • Behavioral and Social Phenomena Important • Deductive Analytic Approaches Too Limited • Computational Power Relatively Unlimited • Convergence • Traditional modeling & simulation • Gaming paradigms & technology • Artificial intelligence • Agent-based modeling
Challenges for ABM • Realistic Models of Agents • Synthetic • Human • Realistic Modeling of “Physics” of Environment • Measurement & Analysis of Behaviors • Inference of New, Emergent Phenomena • Verification and Validation • Models • Behaviors
Needs for ABM • Data Sets • Benchmarks • Standard problems • Methods & Tools • Design • Prototyping • Verification & validation • Testbeds • Physics of environment • Immersive interactive venues
Summary • Complexity & Complex Systems • Definitions • Characteristics • Views • Studies of Complexity • Fault Diagnosis • Large-Scale Systems • Disease Control • Workshop on Complex Systems • Motivation • Agenda • Results • Recommendations • Agent-Based Modeling • Opportunities • Challenges • Needs