1/15 Slide 1KES-B Project Final Presentation – ESRIN, Frascatti, 6th April 2005 1/15 Slide 1KES for DSS in Control Rooms – CESADS(KES) Case Study at ESA/ESOC.

Slide 1KES-B Project Final Presentation – ESRIN, Frascatti, 6th April 2005Slide 1KES for DSS in Control Rooms – CESADS(KES) Case Study at ESA/ESOC

Knowledge Enabled Services (KES)for

Decision Support in Control Rooms.

The CESADS(KES) Case Study at ESA/ESOC.

FR1.3-5I Javier Busto, GTD Sistemas de Información

iCALEPCS 2005 ConferenceGeneva, 10-14 Oct 2005

gtd SISTEMAS DE INFORMACIÓN


Outline

• Context• Addressed Problem• Methods & Solution Approach• Results• Conclusions


Context The European Space Agency (ESA), from its Space Operations Center (ESOC) located in Darmstad (Germany), is able to operate up to 15 Spacecraft (S/C) missions in paralel, using a worldwide network of Ground Stations (G/S) called the European Space Tracking Network, ESTRACK.

ESA/ESOCSpace Operations Center

(Darmstad, Germany)

Kiruna(Sweden)

Redu(Belgium)

Kourou(French Guiana) Villafranca

(Spain)

Perth(Australia)Malindi

(Kenya)

ESTRACK

New missions, specially low-orbit ones like ENVISAT, with high frequency of pass over the G/S lead to critical and knowledge intensive tasks to the mission control operators, posing a stressed and limiting scenario.

During S/C Mission Control lifecycle, one of the most impacting problems is that of the Space Link (S/L) loss.


Outline



Adressed Problem• In particular: to build a Centralised ESTRACK Anomaly and

Diagnostic System (CESADS), to support ESOC/ESTRACK operators in its tasks.

• CESADS goals:– End-to-End Space Link (S/L) Monitoring and Diagnostics– Integrated Supervision of Independent Systems Status

• Mission Control System (MCS), • G/S Monitoring & Control System (CSMC), • Network Management System (NMS), • Flight Dynamics (FOCC).

– S/L Knowledge Management: Capture, Maintain, and Apply the Knowledge of the experts.

– Reduce Operator Stress by providing with integrated monitoring, diagnostic and troubleshooting guidelines.

• In general: the problem of Decision Support in Control Rooms, and the Knowledge Enabled Services (KES) approach.


Outline



Methods & Solution Approach• Development Methodologies:

– Knowledge Engineering: CommonsKADS– Software Engineering Agile methods: Rational, DSDM.

• Knowledge Technologies applied:– For K.Representation (K.R.):

• For Domain Knowledge: Frame based models (leading to Ontologies)• For Inference Knowledge: Declarative Production Rules• For Uncertainty handling: Fuzzy Logic

– For K. Acquisition (K.Acq.)• Manually, with domain experts supported by a forms-based K.Acq.Tool.• Automatic learning, on the basis of enough historical log data and operators

feedback over the generated alarms.– For K.Application (K.App)

• Fuzzy Rule Base Reasoning Engines

• SW Tools & Architectures:– Protégé: Frame&Ontology modelling & KB system platform.– Jess & FuzzyJess: Reasoning engine– Java2 Enterprise Edition (J2EE) platform– RMI-based Observable-Observer pattern in Client-Server

communications

CommonKADS


Development Methodology: Combination CommonKADS + SW Agile Processes (RUP , DSDM)

RUPSystem

Development

RUPSystem

Development

RUPSystem

Development

cKADSKnowledge Management Knowledge Engineering

cKADSKnowledge Management Knowledge Engineering

Organisation ModelTask ModelAgent Model

Organisation ModelTask ModelAgent Model

Communication Model

Communication Model

Knowledge Model

Knowledge Model

task(s) selected in feasibility study and

further detailed in Task and Agent Models

knowledge-intensive task(s)

DesignModelDesignModel

requirements specification for interaction (control) functions

requirements specification for reasoning functions

Context Analysis Concept Analysis Artifact Analysis


Knowledge-intensive Task Types

CommonKADS classifies the knowledge-intensive tasks in two main groups:

• Analytic Task– Classification– Assessment– Diagnosis– Monitoring– Prediction

• Synthetic Task– Design– Modelling– Planning– Schedulling– Assignment

For each one, cKADS provides reference Knowledge Model Templates.


DiagnosticMonitoring

Template for Monitoring & Diagnosis tasksAdapted to knowledge techniques applied (fuzzy logic).

diagnosis

Verify

Additionalobservation

abstractionknowledge

Abstract

TransferFunction

DomainConcept

Inference

Imported crispy variables

fuzzyfication

InitialObservation

abstractionknowledgeAbstract

Import

Initialparameters

Imported variables

In fuzzy terms

Fuzzy Sets

Abnormalityobservation

Normalityobservation

behaviourknowledge

Detect

Fuzzy inference

Indicators

hypothesis

causeknowledge

Hypothesize

Possible causes for a problem(vector of hypothsis)

Additionalparameters

associationknowledge Specify

Import

What can verify the hypothesis?


Design Model: Pattern of a Knowledge Based System

A typical KBS is composed of three main elements:

• A Knowledge Acquisition Tool,

• A Knowledge Base, having Domain Knowledge and Inference Knowledge,

• A set of Problem Solving Methods (PSM), having appropriate Inference Engines, and effectively operating over the Aplication Domain (operational) data.

ApplicationOperationDatabase

Problem Solving Methods (PSM)

InferenceEngines

InferenceEngines

ProductionRule engine

Experts

Operate over

CBRengine

DomainKnowledge

Structure Behaviour

InferenceKnowledge

ProductionRules

CaseBase

Knowledge Base (KB)

Operators

Jess Tool

Monitoring Tool

load

Knowledge Acquisition Tool(based on Protege)

Operation Console


Development Phases: 3 iterations following the DSDM Agile Methodology approach.

CESADSConcept

Design Prototype:Single-mission (Envisat)Simulated systems interface (CSMC, MCS, NMS, etc)Integrated MonitoringRule-Based Inference

Operational Prototype:Single-mission (Envisat)Real systems interface (CSMC, MCS, NMS, etc)Integrated MonitoringKnowledge ManagementRule-Based InferenceBasic Diagnosis

CESADS Concept: Multi-mission Multiple Systems Interface Integrated Monitoring Enhanced Knowledge Mngt. Enhanced Diagnosis Variety of Inference Methods Configurable Scalable

CESADS Project Scope

Functional Prototype

Scope & feasibilityTechs demonstration.

Functional Prototype:Knowledge ElicitationcKADS Knowledge ModelProblem Solving Methods prototype, Inference Engines

Design PrototypeSW design & development

Operational Prototype

Integration in contextTests&Validation


Outline

• Context• Addressed Problem• Methods & Solution Approach• Results: The CESADS Testbed

Prototype• Conclusions


Results: The CESADS Testbed Prototype System

ESOC

G/S

ESA InternalLAN

ESOC SystemsESOC

Systems

CSLAM

!

CMAT

CKAE

CRET

CESADS SystemCESADS System

CESADS Clients

CESADS Clients

CESADS S/L Anomaly and Monitoring Tool

CESADS Management Tool

CESADS Knowledge Acquisition Tool

CESADS Reporting Tool

CESADS Server

CESADS Server


-min_value : value_type-max_value : value_type-frequency-units

Variable

FuzzySet

-status-name : String-description : String

Element

1

*

-status-gradedValues-expert_help : String-firedCounter : Integer-OKfired : Integer-KOfired : Integer

Indicator

-name : String-description : String-applicability : operational_phases-sampling_rate : Integer-status : status

Process

-input_parameters : Parameter-output_parameters : Indicator

Inference Process

-ip_source : String-path_source : String-output_parameters : Variable

Import Process11..*

1

*

-status-name : String-description : String-value_type-rawValue : value_type-fuzzyValue : FuzzySet-generator : Process

System Parameter

Is generated

1 1

-knowledge base

Rule Set

-inference1

-static knowledge1

-ruleID

Rule1

1

Is generated

1*

Knowledge Model: Behaviour & Workflows •Elements (CSMC, SEVT, NMS…)•System Parameters (Variables and Indicators)•Variables (throughput, delay, TM_Counter…)•Indicators (TC_Ind, throughput_Indicator, Network_Status…)•Rule Sets: System

ParameterIndicator

systembehaviour

indicates

1..* 1


C-SLAM: S/L Monitoring & Analyis Client tool Synoptic View

Event Log.

Time and Events.Start-End of the Pass:AOS: Acquisition of Signal.LOS: Loss of Signal

ESTRACK Top-Level Indicators


C-SLAM: S/L Monitoring & Analyis Client tool Indicators Table View, with Alarm System Explanation

Form to acquire formal (ontology-enabled) feedback from the Operator.

Fired Rules that produced the alarm, and input Parameters originating that firing.

Table of indicators generated by the system.


C-KAE: Knowledge Acquisition Tool (1/8) Domain Abstraction Knowledge: Vocabulary of Fuzzy Sets & Fuzzy Terms

Catalogue of FuzzySets for the abstraction in (Fuzzy Terms) of the Parameters (Variables&Indicators)

Enabling the mapping of quantitative values of parameters, into qualitative fuzzy terms (fuzzyfication).

E.g.: Fuzzy Terms for networkLatency concept.


C-KAE: Knowledge Acquisition Tool (2/8): Domain Structure Knowledge: Hierarchy of Elements & subElements in the domain.

4 main subElements compose ESTRACK

Modelling the hierarchy of Elements and its associated Parameters (external imported Variables and inferenced Indicators).

6 main (highlevel) indicators have been define to monitor ESTRACK.


C-KAE: Knowledge Acquisition Tool (3/8) : Domain Structure Knowledge: Paremeters associated to the Elements

Fuzzy Sets selected to represent the values of the parameters.

Modeling the Parameters, using the fuzzy sets abstraction knowledge.

Settings colors for display.


C-KAE: Knowledge Acquisition Tool (4/8) : Domain Behaviour Knowledge: ProcessesModeling the unitary Processes (Import or Transformation/Reasoning) that can be executed by the system.

Highest level SpaceLinkFuzzyInferenceSystem process.

Processes defined:7 Data Import processes6 Fuzzy Inference Processes


C-KAE: Knowledge Acquisition Tool (5/8): Domain Behaviour Knowledge: Workflows

Two main workflows:During PrePass-PhaseDuring Pass-Phase

Defining the workflows of processes, to be executed during the different operational phases.

Processes to be executed by this workflow.

Sub-Workflows can be chained.

Parent workflow.


Rule edition panel.

C-KAE: Knowledge Acquisition Tool (6/8) : Behaviour (Inference) Knowledge: Rulesets of the Fuzzy Inference Processes.

Rule sets.


C-KAE: Knowledge Acquisition Tool (7/8) : Validation of Knowledge Consistency: Protégé Axiom Language (PAL) rulesChecking that Domain Knowledge settings and Inference Rules are consistent.

Consistency rule description and code.

PAL is analogue to what represents be OCL in a UML model.

Protege allows to define and execute Constraint rules over the Frame based model.


All Acquired Knowledge is Stored in an XML-based schemas, and loaded into the CESADS-Server system, in a SQL database.

When the C-Server is started, it creates all the Processes Threads agents, and starts the execution of the controlling workflows.

Then, it is ready to serve the Parameter Objects to which C-SLAM monitoring& diagnostic client application can subscribe to.

C-KAE: Knowledge Acquisition Tool (8/8) : Storing and loading Knowledge Model into the CESADS-Server.


Outline



Conclusions• CommonKADS has been proved as a valid Knowledge Management and Engineering

methodology.

• Knowledge technologies provide added value to the CESADS Adressed Problems:– Knowledge Formalisation: Frame Based systems, Ontologies.– Rule Based Inference Production Engine– Fuzzy Logic Rules.

Successful generation of high-level qualitative indicators, providing an integrated view of a big system (ESTRACK).

• Evolution: however, more AI technologies (particularly that of the Ontologies) are arriving to the IT mainstream (SemanticWeb vision), and therefore a deep platform evolution is envisioned, towards a concept of:

“Knowledge Enabled Services (KES) Grid for Decision Support”.

… as follows:


XMLWeb

InternetGrounding, Foundations

Enterprise PlatformsTechnology

J2EE .NET

WebServicesTechnology

BPEL

SOAPWSDLUDDI ebXML

Technology Umbrellas

TechnologyStandards

Technology Components

E-Gov Network Centric

Operations(NCO)

E-Science AmbientInteligence

E-BusinessE-CommerceTargetted

Needs

ECMCRMERPBI

Typical Business Solutions

SemanticTechnology

RDFOWL

Web-based “Knowledge Enabled Services (KES) Grid” PlatformWebServices Business Process Mngt.Semantic Web Grid Computing

GridTechnology

OGSA

WS-IWS-RF

Distributed Computing

CollaborativeEnvironment

GlobusToolkit-4

Middleware& integration

ApplicationServers

EAIESB

SWS (OWL-S, WSMO)

ServiceRegistries

ServiceOrchestration

SOA

Ontologies

Metadata &Semantic ContentAnnotation

Information Exploitation chain

Knowledge Represent.& Sharing

Conclusion in terms of SW Evolution for KES platformsConvergence of major tech branches into a “KES-Grid” platform concept.Key role of Semantic Web & Ontologies, and particularly the W3C´s RDF/OWL-XML languages, as enabler of a semantic interoperability layer, and sharable knowledge representation.

Frameworks

Academy Institutional IndustryDefense&Security

PrivatePublicResearch CommercialStakeholders

RDF&OWL are being integrated as specific application components, but also enabling extension of other web-based communication and control protocols.


• CommonKADS approaches (Knowledge Intensive Tasks analysis, Knowledge Models Templates, Libraries of Problem Solving Methods) can be migrated to these new state-or-art technology landscape:– Domain Knowledge to be represented with RDF/OWL ontologies.– Domain Behaviour as workflows based on WS-BPEL, chaining distribued &

loosely coupled inference engines that shall be view as webservices.– Domain Elements that are provide useful Parameters for the Monitoring and

Diagnostic task, can publish parameters semantically annotated (e.g., RSS feeds)

• Used tools (Protégé, Jena) support this migration:– Protégé is most-widely use ontology modelling tool.– It integrates Jena (from HP Laps) to support OWL & RDF– JESS is evolving, last version now with an IDE based on Eclipse plugin.

• Thus going towards the combination of benefits of the well founded cKADS Knowledge Engineering methodology, toguether with the emerging benefits of the Semantic Service Oriented Architectures.

Conclusion in terms of SW Evolution for KES platformsCommonKADS approaches over a KES-Grid platform concept.


Questions and Answers

Thank You

Javier Busto, GTD Sistemas de Información, Email: [email protected];

Javier Varas, GTD Sistemas de Información, Email: [email protected] ;

Contacts:

1/15 Slide 1KES-B Project Final Presentation – ESRIN, Frascatti, 6th April 2005 1/15 Slide 1KES for DSS in Control Rooms – CESADS(KES) Case Study at ESA/ESOC.

Documents

informacin slide

esaesoc context

mission control operators

esaesoc outline context

esaesoc adressed problem

sc mission control lifecycle

knowledge engineering

sl knowledge management