methodologies for understanding behavior of system of systems cihan h dagli, phd professor of...

Methodologies for Understanding Behavior of System of Systems

Cihan H Dagli, PhDProfessor of Systems Engineering, Computer

Engineering and Engineering Management

Systems Engineering Graduate Program [email protected]

University of Missouri – RollaRolla, Missouri, U.S.A.

.

January 24. 2007 INCOSE Midwest Gateway Chapter Presentation

2

Understanding Behavior of System of Systems

Are there solutions in the nature?


3

How the nature does it?


4



5



6

Dumb parts, properly connected

into a swarm, yield smart results



7



8

“ Bugs”Can they be the

solution?



9



10



11

KISS …Keep It Simple Stupid



12



13

Most biological systems do not forecast or schedule They respond to their

environment — quickly, robustly, and adaptively

As engineers, let us don’t try and control the system.. Design the system so that it controls and adapts itself to

the environment



14



15

Outline

Introduction System of Systems (SoS) characteristics Network centric operations Challenges and research need for SoS analysis The purpose of the presentation

Modeling Tools for Understanding Behavior of SoS Methodology customization for SoS analysis Conclusions


16

Introduction

We are increasingly a networked society: Trans-national mega military systems Asymmetrical threats vs. rapid reaction forces Trans-national enterprises Globally distributed services and production

We are increasingly dependent on these networks.

Global Net-Centric Operations


17

Evolution of FORCEnet

2020+20092003

Future Vision

AirAir

Today

Seamless Integration

Maritime

Ground

Air


18

Boeing 787 Example

Super-Efficient , Eco-Friendly, and People Friendly

Systems need to be designed for fuzzy attributes


19

Boeing 787 Example

Net-Centric Manufacturing


20

Boeing 787 Example

Fixed and movable leading edges, flight deck, part of forward fuselage, engine pylons

Kansas, Oklahoma Boeing Wichita (announced Nov. 2003; April 2004)

Vertical tail assembly, movable trailing edges, wing-to-body fairing, interiors

Washington, Canada, Australia

Boeing Fabrication (announced Nov. 2003)

Horizontal stabilizer, center fuselage, aft fuselage

Italy, Texas Alenia/Vought Aircraft Industries (announced Nov. 2003)

Airplane development, integration, final assembly, program leadership

Washington Boeing Commercial Airplanes (announced Nov. and Dec. 2003)

787 Work Statement Main Location Company/Business Unit



21

Boeing 787 Example

Auxiliary power unit, environmental control system, remote power distribution units, electrical power generating and start system, primary power distribution, nitrogen generation, ram air turbine emergency power system, electric motor hydraulic pump subsystem

Connecticut Hamilton Sundstrand (announced Feb. 2004, March 2004, July 2004, Sep. 2004)

Wing box Japan Mitsubishi Heavy Industries (announced Nov. 2003)

Main landing gear wheel well,main wing fixed trailing edge,

part of forward fuselage

Japan Kawasaki Heavy Industries (announced Nov. 2003)

Center wing box, integration of the center wing box with the main landing gear wheel well

Japan Fuji Heavy Industries

(announced Nov. 2003)




22

Boeing 787 Example

Fuel quantity indicating system, nacelles, proximity sensing system, electric brakes, exterior lighting, cargo handling system

North Carolina Goodrich ( announced March 2004; April 2004, June 2004, Nov. 2004, Dec. 2004)

Common core system, landing gear actuation and control system, high lift actuation system

United Kingdom Smiths (announced Feb. 2004, Jun. 2004)

Navigation, maintenance/crew information systems, flight control electronics; exterior lighting

Arizona Honeywell (announced Feb. 2004, July 2004)

Displays, communications/ surveillance systems

Iowa Rockwell Collins (announced Feb. 2004) )




23

Boeing 787 Example

Composite mat for the wing ice

protection system United Kingdom GKN Aerospace (announced

Dec. 2004)

Wireless emergency lighting system

Arizona Securaplane (announced April 2005)

Global collaboration tools/software

France Dassault Systèmes (announced Feb. 2004)

Landing gear structure France Messier-Dowty (announced March 2004)




24

Virtual FunctionalNetworks

Virtual MissionNetworks

Boeing 787 Example

Rolls-Royce

General Electric

Messier-Bugatti

Latecoere

Fuji Heavy Industries

Boeing Commercial Airplanes Alenia/Vought Aircraft

42 Global Company Information Grid


25

Complex Systems Architecting

Seamless integration and dynamic adaptation to changing environments are common characteristics.

These characteristics are true for both defense and commercial systems

Resulting systems are complex; their behavior can be understood through Computational Intelligence, Artificial Life approaches and Complexity Theory

They can only be created with evolving architectures


26


Attributes for Complex Systems Interdependent Independent Distributed Cooperative Competitive Adaptive

Attributes defines the complex system


27


These attributes are being used to create new system definitions Systems of Systems (Interdependent) Family of Systems (Independent) Galaxies of Systems (Distributed) Intelligent Enterprise Systems (Cooperative,

Competitive and Adaptive)

Recent systems definitions can be based on attributes


28

Characteristics of SoS

System of systems (SoS) – A set or arrangements of interdependent systems that are related or connected to provide a given capability.

Common Characteristics: Operational independence of

elements Elements possess the

required NCO interoperability Development and existence is

evolutionary Emergent behaviors and

capabilities Geographically distributed

NCO systems

Interdependent complex system


29

Network Centric Operations

Effects based planning Effects based operations Global reach Information superiority Collaborative decision

superiority Horizontal and vertical

integration Joint deployment and

sustainability Net centric coalition

Comprehensive situational awareness

Common Operating Picture Rapid, tailored response

and agility Supports layered approach

to security Embedded simulation and

training

Attributes of NCO architecture as a complex system


30

GIG Complex Systems

All of these systems need an physical global interface to function

A Global Information Grid represents the system formed by the distributed collections of electronic capabilities that are managed and coordinated to support some sort of enterprise (virtual organization). Traditional, large complex system A service-oriented architecture (SOA) is proposed by

Berman, Fox and Hey SOA is essentially a collection of services which

communicate with each other. The communication can involve either simple data

passing or it could involve two or more services coordinating some activity.

Global Physical Architecture is a Complex System


31

G I G

GIG Complex System

An Evolving Global Physical Architecture


32

Net-Centric System

An Evolving Net-Centric Architecture

G I G

Net-Centric Architecture

Robust

Interoperable

Adaptable

Flexible Modular


33

G I G

Net-Centric Architecture

Robust

Interoperable

Adaptable

Flexible Modular


System 1

Meta-Architecture

Dynamically Changing Meta-Architecture for Complex Systems

System 2

System 3

System 4

System n

System n-1


34


GIG and dynamic net-centric architecture provides the basic interface for creating meta-architectures of the complex systems of this century.

It is the dynamically changing architecture that creates the best net-centric systems not the data that passes through it, although it is a necessity for the system to function.

Systems Architecting and Complexity Theory are essential in designing net-centric systems


35

Challenges and Research need for SoS

Dynamically changing requirements increase uncertainty.

Continuous rapid technological changes provide opportunities for improved capabilities, but increase complexity of interfaces.

Diverse spectrum of missions and operations increase complexity of architecting SoSs.

The need to develop dynamic and evolving communication architecture is vital in architecting SoS.

Adopting to the environment is must in system architecting


36

Challenges and Research need for SoS

Interoperability is the main challenge which must be met. Interoperability-related enablers, such as

Architecture frameworks Technical architectures Levels of information systems interoperability

The challenge arises due to missing linkages between these interoperability-related enablers and SE processes.

All these challenges open various research needs for SoS

Interoperability helps to create meta architectures that can adapt


37

The Purpose of the Presentation

Discuss several computational intelligence tools for modeling and understanding behavior of SoS.

Identify the areas where these tools can help SoS architects in dealing with the mentioned challenges.

Identify methodology customization for analysis of SoS.

Are there approaches that we can use for understanding behavior of complex systems?


38

SoS Conceptual Framework

Robust Physical

Networks

Robust Information

Networks

Robust SocialNetworks:

-PeopleOrganizations

-Processes

Better networking and information sharing

Improved situation awareness/understanding

Enhanced collaboration/interactions

More agile SoS elements

Improved effectiveness

Diversity and robustness can help


39

SoS Design Approach

Requirements Generation &

Mission Analysis

Analysis of Alternatives, Technology Insertion & Collaborative

Innovation Plan

Systems of Systems Engineering, Plan Synthesis and Risk

Assessment

Acquisition Planning, Test & Evaluation & Configuration

Control

Spiral feedback

Computational intelligence




Computational Intelligence Approaches can Improve Architectures


40

Modeling Tools for Understanding Behavior of SoS

Distributed systems modeling Complexity theory Agent based modeling Evolutionary strategies and programming Swarm intelligence and optimization Emergent behavior analysis of architectures Fuzzy logic Cooperative and collaborative system modeling

and simulation Adaptive system architecture generation

Architect’s Tool Box


41

Formulation of Distributed Networked Systems

Distributed System Building

Blocks

Distributed System

Functionality

Defining Distributed

System Characteristics

Distributed System Design

Goals

Distributed System Design

Principles

Engineering Distributed

Systems

Distributed Systems Modeling


42

Trajectories of Research into Distributed Systems Modeling

System Behavior &

Analysis

System Behavior & Analysis

System Design

System DesignSwarm

Intelligence & Synthetic

Ecosystems

Artificial Life

Multi-Multi-agent agent

SystemsSystems

Distributed Artificial

Intelligence

Population Biology&

Ecological Modeling

Distributed Systems Modeling Approaches


43

Complexity Theory and SoS analysis

Long term planning is impossible: SoS grow by adding components

Dramatic change can occur unexpectedly: Small perturbations can also cause huge

changes on the overall system behavior Potential for cascading failures in SoS

Is there are room for complexity theory?


44

Complexity Theory and SoS analysis

Complex systems exhibit patterns and short-term predictability: Next time period behavior of systems can be predicted

when reasonable specifications of conditions at one time period are given

SoS testing and validation is based on this characteristic Organizations can be turned to be more innovative and

adaptive: Emergent order and self organization provide a robust

solution for SoS to be successful in competitive and rapidly changing environmental conditions

Is there are room for complexity theory?


45

Agent-based Modeling and Sos Analysis

perceive action

Agents Environment

System

Communication

Rules

update update

receive

transmit

SoS

Sub-Systems

Distributed Information Interfaces

Agent Based Architecture


46

Distributed Agent Paradigm

Cooperate Learn

Autonomous

Collaborative Learning Agents

Smart Agents

Interface agentsCollaborative Agents

Agent Types


47

Evolution in Agent Paradigm: Reinforcement Learning & Genetic Algorithms

Population(Decision Rules)

Match Set

Prediction Array

Action Set

Environment

Detectors Effectors

Previous Action Set

GA P

Delay=1

RewardPerformance Component

Discovery Component

Reinforcement Component

Evolutionary Modeling


48

Swarm Intelligence Systems

Entities Share Common Goal

Local Interaction

sSelf

Organization

Autonomy of Units

Stigmergy

Simple Rules or Units

Distributed

Large Number or

Efficient Size

Pulsing of Force

Flexible and Robust

Swarm Attributes


49

Wasp - Introduction

Biological studies [Wilson 1971, Pratte 1989, Roseler 1991, Reeve and Gamboa 1987, Gamboa et al. 1990, Chandrashekara and Gadagkar 1991]

Based on the above hypotheses, several researchers built mathematical and analytical models of wasps [Theraulaz et al. 1991, Theraulaz et al. 1992, Bonabeau et al. 1996, Robson and Beshers 1997, Page and Mitchell 1998, O’Donnell 1998]

Bonabeau et al. 1996, 1997 – Detailed study of wasp behavior and applied it to practical toy problems

Can we model wasp’s behavior?


50

Wasp - Introduction

3 Level Hierarchy Queen (and Elite Class) Workers Nurses

Force

Threshold

Queen

Workers

Nurses

Two main attributes for division of labor


51

Wasp - Introduction

Dominance Contests Higher force – greater chance of winning Division of labor

Foraging Lower Threshold – more foraging Higher Stimulus – more success ~ High Force

Dominance and Foraging


52

Wasp - Introduction

Stigmergy

High Demand Forage Reduces Demand

Low Food Egg Laying Stay in Nest

Threshold

Threshold

Force

Threshold

Stimulus

Stigmergy as Swarm Attribute


53

Wasp - Introduction

Probability of wasp 1 winning over wasp 2 for a given set of force variables

F2

2

2

1

2

1

FF Wins) 1 (Wasp P

Mathematical Representation of Wasp’s Behavior


54

Wasp - Introduction

Threshold updatesθw = (θw,1, ……. θw,t)

Probability of successful foraging based on stimulus and thresholds

2

,

2

t

2

t

SS task t)a perform wP(Wasp

tw

Mathematical Representation of Wasp’s Behavior


55

Wasp - Introduction

Simple rules Stigmergy Distributed operation Emergence Minimal and indirect communication

Attributes of Wasp’s Behavior


56

Wasps for Manufacturing Systems

Wasps Manufacturing System

GoalTo maximize food

collection, egg laying,nest building

To maximize throughput, minimizenumber of setups incurred,

minimize average cycle time

Agents Wasps Machines

Work Specialization

Wasps specialize to gather food or build nest

or lay eggs

Machines specialize to processparticular part types and avoid

additional setups

ForceForce variable of wasp Remaining processing times, setup

times, wait times of machines

ThresholdThreshold of wasp

Setup requirements

StimulusScent of food

Waiting times of parts

It works better than classical approaches


57

Swarm Systems and SoS Analysis

Military Swarm Scenario

Collaborative Swarm Robots

Swarm Routing in Communication Networks


58

Emergent Behavior Analysis of SoS Architectures

Three step analysis1. Structural approach: Model and develop a common

understanding platform for SoS using Department of Defense Architecture Framework (DoDAF)

Operational View Systems View Technical Standard View

2. Object-oriented approach: Model and identify end users’ requirements, states and sequence of events that system can undergo

3. System behavior analysis: Use an executable model such as Petri-Nets for architecture evaluation and validation


59

Emergent Behavior Analysis of SoS Architectures

Structural Approach

Object-oriented Approach

Executable Model

SoS Architecture

Modify

Modify

Modify

Executable architecture is a must for SoS


60

Areas Computational Intelligence Tools can help SoS architects

Each architecture provides a compromise between four attributes: system cost, schedule, system risk and system performance

As these attributes change based on environment compromises change and make the architectures dynamic rather than static.

Architectures need to evolve in time for SoS


61

Areas Computational Intelligence Tools can help SoS architects

The role of computational intelligence tools is to aid SoS designers in critical technical analysis: Decision aiding algorithms Testing and validating as well as development of

Command and control architecture Communication links Logistic infrastructure Common architectures/modules Sensor technology Platform capabilities

Computational tools can help


62

Methodology customization for SoS analysis

New modeling and simulation algorithms based on biologically inspired approaches should be added to systems engineer's tool box to cope with modeling and analyzing emerging systems.

Ability to learn and evolve new architectures from the previously generated ones, based on systems performance values, need to be incorporated in modeling and simulation process.

System Architect needs new computational intelligence based tools for effective search


63


DoDAF architectural framework should be modified for commercial SoS

Both structural and object-oriented analysis is required for comprehension of SoS

Simulation tools that combine various modeling paradigms (discrete, agent-based, system dynamics) should be used in analysis of SoS to capture different behavioral views

Heuristics are not sufficient to do the job


64


Supervised learning technique cannot handle rapidly evolving SoS.

A supervised learning assisted reinforcement learning architecture is more suitable for modeling data prediction and analysis in SoS.

Adaptability for dynamic architectures can be achieved through learning


65


Therefore, Adaptive Critic Designs and Q-learning will be considered as potential reinforcement learning candidates.

After sufficient coarse learning, fine learning is applied which employs reinforcement learning algorithms.

Reinforcement Learning agent will model system evolution

Reinforcement learning is an alternative


66

Conclusions

Modeling and simulation for generating architecture alternatives is essential.

Classical systems engineering practices, modeling and simulation approaches need to evolve to cope with system of systems.

New engineering tools are required to complement the existing ones for modeling and simulation and automatic architecture generation.


67

Conclusions

Computational intelligence tools help in creating the desired behavior for the SoS


68

Most biological systems do not forecast or schedule They respond to their

environment — quickly, robustly, and adaptively

As engineers, let us don’t try and control the system.. Design the system so that it controls and adapts itself to

the environment



69

Are we there yet?

Thank You

Thank You

methodologies for understanding behavior of system of systems cihan h dagli, phd professor of...

Documents

computer engineering

missouri rollarolla

research need

networked society

systemscihan h dagli

presentationmodeling

distributed services

simple stupid