Upload
kelly-parsons
View
217
Download
1
Tags:
Embed Size (px)
Citation preview
TOPIC 9: SYSTEM ENGINEERING
David L. Hall
TOPIC OBJECTIVES
Discuss issues associated with the definition, design and implementation of a data fusion system
THE IMPLEMENTATION ENVIRONMENT
Combined Hardware and Software DevelopmentCombined Hardware and Software Development Procured commercial hardware Specialized hardware Large-scale, real-time software
Complex System ArchitectureComplex System Architecture Multiprocesssing Distributed database and processing External interfaces
Stringent (and Changing) RequirementsStringent (and Changing) Requirements Throughput/response requirements Platform constraints (air/land/sea) Deployment environment Reliability/availability
Multiple Users and DevelopersMultiple Users and Developers Security RequirementsSecurity Requirements
Multi-level securityMulti-level security
Enterprise Services EnvironmentEnterprise Services Environment Service Oriented Architectures (SOA)Service Oriented Architectures (SOA)
FUSION IN THE CONTEXT OF REAL SYSTEMS
Type N Sensor
• Obs Modeling• Resource
Optimization• Goal Prorgamming
• Obs Correlation• Parameter Estimation• Pattern Recognition
Sensor Management
Single Sensor Obs Processing
Single Sensor Obs Processing
Multiple Sensor Obs Processing
Decision Aid/Planning Support
• Situation Assessment• Planning
• Multiple Sensor Correlation
• Parameter Estimation• Pattern Recognition
HCI
Analyst
Sensors
• Input Message - Event Detection - Alerts - Routing• Output Message - Routing
Me
ss
ag
e
Pro
ce
ss
ingData
Processing
• Smoothing• Preliminary Target I/D
SensorTasking
Co
mm
un
ica
tio
ns
Dat
a B
ase
Man
agem
ent
Analyst. . .
PROFILE OF DATA FUSION SOFTWARE FUNCTIONS
Dat
a F
usi
on
Dat
a P
roce
ssin
g
Dat
a M
an
agem
ent
OP
Sy
stem
Co
mm
un
icat
ion
s
Har
dw
are
Co
ntr
ol
S
pec
ial
Dis
pla
y
Use
r In
terf
ace
Pro
cess
ing
Co
ntr
ol
15
10
5
20
Per
Cen
t o
f T
ota
l S
oft
war
eP
er C
ent
of
To
tal
So
ftw
are
Categories of Software Functions
REQUIREMENTS FLOWDOWN PROCESS FOR SENSORS
REQUIREMENTS FLOWDOWN PROCESS FOR DATA FUSION
Sensor Performance Requirements
Data Fusion Systems
Requirements
Subsystem Design
Process
Processing System Design
Sensor Systems Design
• Coverage
• Detection
• Tracking
• Classification
• Algorithms
• Database
• Architecture
Display System Design
• HCI
Communications Design
• Capacity
• Architecture
Algorithm Requirements
Process Requirements
Display Requirements
Sensor Requirements Communication
Requirements
Performance Analysis
Simulations
SCENARIOS
Test/Evaluation Requirements
DISTRIBUTED FUSION ARCHITECTURES
F
C
F
F
F
C
C
F
F
F
F
C
C
F
C
Sensor Node
Fusion NodeInformation Consumer Node
Centralized Hierarchical Distributed
SS
S
S
S
S
S
S
S
SS S
S
S
S
S
S
S
S
Sensor Data Flow
Knowledge Level Info. Flow
Inference Message Flow
Courtesy, M. Liggins
INFERENCE ARCHITECTURES – CENTRALIZED
Inference Link(association or causal link)
C
S
S
S
S
S
S
-Inference NodeReceive Data from Sensors
Fusion Node
C - Output/query Node
F
CS
S
S
S
S
S
GENERIC TRACKER/CORRELATOR ARCHITECTURES: CENTRALIZED
FUSION
SensorA
SensorB
SensorN
Preprocessing
Preprocessing
Preprocessing
•••
Sensor Controls
Gating and Control Parameters
Target State
Processed Sensor
Data
• Target Classification• Probability of Successful Declaration
CorrelationData Alignment and Association
Composite Filtering
Classification
Level 1 Fusion
GENERIC TRACKER/CORRELATOR ARCHITECTURES: AUTONOMOUS (OR
DISTRIBUTED) FUSION
Sensor Controls
Gating and Control Parameters
Target State
Tracking and Classification Parameters • Target Classification
• Probability of Successful Declaration
CorrelationData Alignment and Association
Composite Filtering
Classification
Detection and Estimation
Tracking & Classification
Tracking & Classification
Tracking & Classification
SensorA
SensorB
SensorN
Preprocessing
•••
Preprocessing
Preprocessing
GENERIC TRACKER/CORRELATOR ARCHITECTURES: HYBRID FUSION
Select and MergeMUX
Sensor Controls
Gating and Control Parameters
Target State
Tracking and Classification Parameters • Target Classification
• Probability of Successful Declaration
CorrelationData Alignment and Association
Composite Filtering
Classification
Detection and Estimation
Tracking & Classification
Tracking & Classification
Tracking & Classification
SensorA
SensorB
SensorN
Preprocessing
•••
Preprocessing
Preprocessing Detection Parameters
SUMMARY OF ALTERNATIVE FUSION ARCHITECTURES
• Minimum information loss• Requires commensurate
sensors• Difficult association problem• Much communication between
sensors/fusion system
ARCHITECTURE TYPE COMMENTSLEVEL OF
INFORMATION FUSEDAPPLICABLE TECHNIQUES
Centralized RAW DATA - direct fusion of sensor data:- Sample signal- Imagery
• Physical models• Pattern recognition• Estimation techniques
• Information loss due to feature extraction
• Large dimension composition feature vector
• Simplified association• Allows non-commensurate
sensors• Reduced communications
• Significant information loss• Local optimization may
prohibit global optimization• Allows non-commensurate
sensors• Minimum communication
Must account for statistical interdependence
• Combines features of centralized and autonomous architectures
• Complex control logic• Increased communications
requirement
Distributed/Autonomous
Hybrid
FEATURE VECTORS:- Attribute features- Locational/ kinematic parameters
DECISION-LEVEL INFORMATION:
- State vectors- Identity
declarations
COMBINATION OF:- Raw data- Feature vectors- Decision level
• Pattern recognition• Estimation techniques• Cluster analysis• Neural networks• Parametric templates
• Estimation techniques• Bayesian inference• Dempster-Shafer's
method• Logical templates• Voting methods
• All of above techniques
Fusion of raw data from sensors
Fusion of derived data (features) from sensors
Fusion of state vectors or identity declarations
Combination of centralized and autonomous architectures
DESCRIPTION
TEST AND EVALUATION ISSUES/COMPLEXITIES
Numeric/Symbolic Components - Testing of Algorithms and Knowledge Acid Tests - with/without fusion
Dominant sensor (1,2,3, … N) source effects/marginal gain Comparison between fusion techniques
Measuring Global Optimality Standards/expectations for optimality (detection level versus mission-level) Complex role of feedback in optimization Broad range of architectural choices (e.g., pixel fusion versus decision level)
Stochastic Details Validity of simplifying assumptions Often interested in rare events (e.g., situations)
No Gold Standard in Situation and Threat Assessment (What constitutes the quality of a situation assessment?)
FOUR CATEGORIES OF MEASURES OF MERIT
MEASURE DEFINITION TYPICAL EXAMPLES Measure of Force Effectiveness (MOFE) Measure of how a C3 system and the force of
which it is a part perform military mission (sensors, weapons, C3 system)
Outcome of battle Cost of system Survivability Attrition rate Exchange ratio Weapons on targets
Measures of Effectiveness (MOE) Measure of how a C3 system performs its functions within an operation environment
Target nomination rate Timeliness of information Accuracy of information Warning time Target leakage Countermeasure immunity Communication survivability
Measures of Performance (MOP) Measures closely related to dimensional parameters (both physical and structural), but attributed to behavior
Detection probability False alarm rate Location estimation accuracy Identification range Time from detection to transmission Communication time delay Sensor spatial coverage Target classification accuracy
Dimensional Parameters The properties or characteristics inherent in the physical entities whose values determine system behavior and the structure under question, even when not operating
Signal-to-noise ratio Operations per second Aperture dimensions Bit error rates Resolution Sample rates Anti-jamming margins Cost
Courtesy of J. Llinas
EXAMPLE OF A FUSION EVALUATION TEST BED
(MITRE CORPORATION)
Post Processing Data Analysis Tools
• Graphs• Statistical Tests between Algorithms
Sensor Models Module
• RADAR• IFF• ESM• Hypothetical Sensor A• Hypothetical Sensor B
Measures of Effectiveness
Output Module
• Track Kinematics• Track Maintenance• Correlation• Classification Identification
Fusion Algorithm Under Test
Live Recorded Data Module
Scenario Generation Module
• Target Files• Platform Files• Scenario Generation File• Administrative Data File
Target/Platform Motion
and Dynamics
Module
Files
Files
DATA FUSION THRUPUT PARAMETRICS
ASSUMPTIONS: • Single-level Tracking• Fixed Processing Loop per Input Report
N = ABCD
R
MOPS
D
C
B
A
H2 Hypothesis
S Searches/Tgt
J MOPS/UP
X
Data Association
Tracking
Overhead
Track File Data Base
Management
Data Combination
Sensor Management
Backward Logical
Inferences
R
H/RM Sensors
N Targets
S1 Visits/Sec
S2 Visits/Sec
•
•
S5 Visits/Sec
Revisit Model
Target Density Model
MOPS+
PARAMETRIC FACTORS IN FUSION PROCESSING
APPLICATION PARAMETERS• Number of Sensor Contributors• Sensor Spatial Resolution• Measurement Bandwidth/
Preprocessing• Search Volume• Target Density• Track/ID Response Time• Scenario Force Mix• Number of Target Classes• Feature/Cond. Probability Variance
ALGORITHM FACTORS• Correlation Track Generalization and
Cluster Approach• Degree of Multi-scan and
Recursiveness• Parametric/Non-parametric
PROCESSING PARAMETERS
• Track Storage
• Dimensionalityof decision space
• Decision Rate(target decisions/sec)
• Decision Rule Storage
(parametric/non-parametric)
• Tracking Recursion
• Track File Interaction
• Decision Rule Test Rate
COMPUTATION PARAMETERS
• Correlation/Classification Arithmetic Operation Rate (MIPS/MOPS)
• I/O Transfer Rate
• Track File Capacity
• Classifier Database Capacity/Access
OPERATIONS ANALYSIS: TRANSACTION ANALYSIS
Sensor 1
Sensor 2• • •
Human Interface
Transaction A
Transaction B
Transaction N
• • • S
yste
m R
eso
urc
e A
nal
ysis
Sys
tem
Re
sou
rce
An
alys
is
• Resource Requirements• Capacity• Response Time
• • •
Analysis Results
• MIPS/MOPS• Transmission BWs• Access Speeds
• • •
System Resources
Ra
tes
TransactionsA B N
• • •
EXPLOITATION OF PARALLELISM
EVOLUTION OF COMPUTING POWER
http://www.en.wikipedia.org/wiki/Moore’s_law
EXAMPLES OF DATA INVOLVED IN MSDF
DATA DATA DESCRIPTION INTERPRETATIONModel Parameters Sensor characteristics
Sensor locations Physical constants Time delays, biases, and model
coefficients
Characteristics modelinformation for observationprediction, state vectorpropagation, and sensor dataprocessing
Sensor Data Observations- State vectors- Attribute/feature vectors- Raw signal or imagery
Observed data from sensors
External Databases Previous observations Data from external sensors or
fusion systems
Related data from other fusionsystems or external databases
Human Input Control information Inferences/hypotheses Requests for data Annotations
Data input by humans forsystem control or data update
Environmental Data Geography/topology/hydrology Weather Environmental conditions Man-made objects (e.g., cities,
roads, and networks)
Information about the static ordynamic environment in whicha situation is observed
EXAMPLES OF DATA INVOLVED IN MSDF
DATA DATA DESCRIPTION INTERPRETATION Situation Database (Level 2)
Location, identity of entities Relationships among entities (physical,
communications, hierarchy, sequential, etc.) Interpretation of order-of-battle
Results of Level 2 processing to achieve a dynamic interpreta-tion of entity relationships, activities, or events
Threat Database (Level 3)
Location, identity of threatening conditions or entities, sensors
Characteristics of threatening conditions (threat envelopes and sensors)
Probable courses of action Interpretation of possible consequences
(intent, lethality, and opportunity)
Results of Level 3 processing aimed at determining an enemy threat to friendly or neutral forces
Performance Data (Level 4)
Assessment of accuracies of fusion products Measures of performance Measures of effectiveness Objective function Constraints
Data that characterizes and allows control of the dynamic sensor observation and fusion process
A Priori Data Knowledge bases (rules, frames, scripts, etc.) or target and event behaviors, enemy tactics, doctrine
Technical data (e.g., enemy sensor performance, weapon characteristics)
Mission data - Objective and goals - Strategies - Tactics
Data developed by a priori (usually extensive) analysis to provide a basis for knowledge-based fusion processing
PUSH/PULL FOR SERVICE ORIENTED ARCHITECTURES (SOA)
Technology Push Requirements PullSOA
- Rapid evolution of Internet - Commercial practices - Wireless, mobile communications - Smart sensor/processors - Utilization of web-based models and tools - Etc
- Heterogeneous smart sensors - Distributed team environments - New threats requiring multi-force, multi-org. participants - DoD directives
SERVICE ORIENTED ARCHITECTURE
Service-Oriented Architecture – a component model that inter-relates the different functional units of an
application, called services, through well-defined interfaces and contracts between these services
The interface is defined in a neutral manner should be independent of the hardware platform, the operating system, and
the programming language the service is implemented in
This allows services, built on a variety of such systems, to interact with each other in a uniform and universal manner
SERVICE ORIENTED ARCHITECTURE
SecurityServices
MonitoringServices
ServiceRegistries
MessagingServices
DataServices
TransformationServices
Service Enabled InfrastructurePublish
Data and applications available for use, accessible via services. Metadata added to services based on producer’s format.
Service Producer
• Describes content using metadata• Posts metadata in catalogs for discovery• Exposes data and applications as services
Discover
Invoke
Automated search of data services using metadata. Pulls data of interest. Based on producer registered format and definitions, translates into needed structure.
Service Consumer
• Searches metadata catalogs to find data services
• Analyzes metadata search results found• Pulls selected data based on metadata
understanding
SecurityServices
MonitoringServices
ServiceRegistries
MessagingServices
DataServices
TransformationServices
Service Enabled Infrastructure
SecurityServices
MonitoringServices
ServiceRegistries
MessagingServices
MessagingServices
DataServices
DataServices
TransformationServices
Service Enabled InfrastructurePublishPublish
Data and applications available for use, accessible via services. Metadata added to services based on producer’s format.
Service Producer
• Describes content using metadata• Posts metadata in catalogs for discovery• Exposes data and applications as services
Data and applications available for use, accessible via services. Metadata added to services based on producer’s format.
Service Producer
• Describes content using metadata• Posts metadata in catalogs for discovery• Exposes data and applications as services
DiscoverDiscover
InvokeInvoke
Automated search of data services using metadata. Pulls data of interest. Based on producer registered format and definitions, translates into needed structure.
Service Consumer
• Searches metadata catalogs to find data services
• Analyzes metadata search results found• Pulls selected data based on metadata
understanding
Automated search of data services using metadata. Pulls data of interest. Based on producer registered format and definitions, translates into needed structure.
Service Consumer
• Searches metadata catalogs to find data services
• Analyzes metadata search results found• Pulls selected data based on metadata
understanding
(Post) (Find)
(Bind)
From: GiG Enterprise Services Piloting; Rob Walker, DISA April 2004
LocatorServices
Measuring and DisseminatingReal-World Information
Report MgtServices
WeatherServices
Decision Making
Planning and ConductingOperations
AssociationServices
Organizing and ManagingData and Information
Entity MgtServices
WorkspaceServices
Process MgtServices
Organizing and ManagingWorkflow and C2 Processes
WorkflowServices
Resource MgtServices
OverlayServices
OceanographyServices
Sense &RespondServices
AlertServices
NGC2* SupportServices Coordinated
ActivitiesServices
VisualizationServices
FusionServices
Sense Making andBattlespace Understanding
IntelServices
ReadinessServices
ForceProjectionServices
Air/SpaceOperations
Services
Joint Fires& Maneuvers
Services
ExecutiveSummaryServices
ForceProjectionServices
SituationalAwareness
Services
. . .
UserManagement
Services
. . .
. . .
. . .
. . .
. . .
. . .
MCPs
Core C2 SupportTools & Functions
NGC2*
CES* DiscoveryServices
Hosting and enterprise accessto information
MediationServices
MessagingServices
ESM*Services
SecurityServices . . .
C2 CoI*CommonServices
* MCP – Mission Capability Packages NGC2 – Next Generation Command and Control COI – Community of Interest CES – Core Enterprise Services ESM – Enterprise Service Management
HIERARCHY OF ANTICIPATED SERVICES
From: GiG Enterprise Services Piloting; Rob Walker, DISA April 2004
ANTICIPATED ADVANTAGES OF SOA
Leverage Existing Assets ( Don’t throw away existing systems) Support all types or “styles” of integration.
User Interaction Application Connectivity Process Integration Information Integration Build to integrate
Allow for incremental integrations and migration of assets Include a development environment that has the following features
Should be built around the component framework Promote better reuse of modules and systems Allow legacy assets to be migrated to the framework Allow for timely implementation of new technologies
Allow implementation of new models: Specifically new portal-based client models, Grid computing On-demand computing
SOA TRANSITION PERSPECTIVES
Implementation realities – how do we transition from core services to an effective system?
Impact on operations – how do we really operate in an SOA environment?
Maintenance nightmares – will SOA be a maintenance nightmare?
IMPLEMENTATION ISSUES (EXAMPLES)
Development of application-level components & services – if we build it they will come
Amount & level of imposed structure - level of federation versus free market
Demands and changes for legacy systems – your problem versus my problem
Growth & evolution of core services – where do we stop? Development of fundamental standards, ontologies, templates – tower
of Babel Security – multi-level security etc Inclusion - boundaries of services & components
OPERATIONAL ISSUES (EXAMPLES)
Use of dynamic discovery/link capability – how will real users use dynamic link capabilities?
Human in the loop processing – role of human as consumers & suppliers
Hierarchical heterogeneous users – different needs, capabilities & priorities
Evolution & changes to components – introduction of new components & changes
Adjudication of resource demands – contention, control of user/suppliers
Impact on legacy systems Etc.
SYSTEM MANAGEMENT (EXAMPLES)
System configuration management Rules of engagement of component entrance Version thrashing Upgrades
System maintenance Core services Components
Level and type of support services Ownership and responsibilities Etc
THE SECURITY ENGINEERING PROCESS
Operational Objectives
Formal Security Policy and Guidance
System Concept
Formal Requirements
Engineering Analysis/Trade-Offs
Security CONOPS
System Security Policy
Secure Applications
Assurances
Software Security Model
CHECKLIST TO SUCCESSFULLY BUILD A DATA FUSION SYSTEM
Do You Understand the Application?
What is the mission? What decisions/inferences are required? What are the timeframe/rates for decisions? Who are the system users? How do decisions affect the mission? What is the typical situation/threat environment? What is the appropriate MOE/MOP? What are the platform/system constraints? What are the other data sources? Is real data available? Do we have (former) users/analysts available? How does mission environment affect system architectures?
Do You Understand the Inference Process?
What is processing chain from energy detection to inferences?
Do you understand the cognitive process for effective inferences?
Can you characterize effective versus ineffective inferences?
What are the barriers to correct inferences (human limitations, CM, etc.)?
What is the use of negative information? What is the decision environment (stress, decision styles,
doctrine, etc.)? What are the effective/applicable processing techniques?
SOFTWARE TOOLS AND SYSTEMS
http://en.wikipedia.org/wiki/Comparison_of_agent-based_modeling_software (Comparison of agent-based software)
http://www.lionhrtpub.com/orms/surveys/FSS/fss9.html (survey of estimation software tools) http://cqm.cs.mcgill.ca/~godfried/teaching/pr-web.html (large set of links to training materials,
tutorials, and software tools for pattern recognition) http://onlinelibrary.wiley.com/doi/10.1002/widm.24/full (extensive list and comments on software
for data mining, data warehousing, and knowledge discovery tools) http://www.cs.cofc.edu/~manaris/ai-education-repository/expert-systems-tools.html (directory of
software for expert systems and knowledge-based systems) http://www.lionhrtpub.com/orms/surveys/sa/sa-surveymain.html (survey of 125 vendors of statistical
software) C. N. Mutchler, “An investigation of tools/systems for multi-source data fusion at the national air
intelligence center,” Aerospace Technical Report TOR-20001(1758)-0396e ;2001 S. A. H. MacMullen and R. Sherry, “A survey of COTS software for multi-sensor data fusion: what’s
new since Hall and Linn?”, Proceedings of the MSS National Symposium on Sensor and Data Fusion (San Diego, CA 13-15 August, 2002)
TOPIC 9 ASSIGNMENTS
Preview the on-line topic 9 materials Read chapter 10 of Hall and McMullen (2004) Read Bowman and Steinberg (chapter 16) in
Handbook of Multisensor Data Fusion
DATA FUSION TIP OF THE WEEK
Picture: Stuart MacFarlane,London, Daily News, October 1999.
The process of designing and implementing a data fusion system may seem overwhelming and painful – indeed numerous data fusion systems don’t actually fuse anything (they develop infrastructure software and run out of time/funding before implementing the fusion software). On the positive side there are significant resources available to support a successful implementation including COTS software.