TOPIC 13: LEVEL 4

David L. Hall

TOPIC OBJECTIVES

Introduce Level-4 processing (process control) Describe two broad approaches for Level-4

processing Optimization techniques Market based methods

Provide insight into new research areas

LEVEL 4 PROCESSING (PROCESS REFINEMENT)

MOTIVATION/EVOLUTION OF LEVEL-4 PROCESSING

Intelligent sensor/processor systems – rapid advances provide opportunity for increased dynamic tasking and control (e.g., agile sensors, advanced waveforms, look-angle computations, etc)

Distributed, dynamic network systems – increased complexity of network systems with heterogeneous, distributed sensors and processors

Increased bandwidth – new opportunities for strategies for passing data and commands throughout an environment

Service based architectures - network services provide dynamic insertion of new sensors, models, processing services

Complex algorithms – new agile algorithms provide potential for hybrid, parallel, multi-model approaches

JDL Level 4 Process: Process Refinement

Mission Management

Target Prediction

Sensor & Platform Modeling

System Performance Modeling

System Control

Representation of mission objectivesMission constraintsAdjudication between fusion optimization & mission constraint

Target state estimation

Target attribute modeling

Sensor platform modelsSensor characteristicsSignal propagation modelsTarget/sensor signal interaction

Sensor performance modelsAlgorithm performance modelsMeasures of performanceMeasures of effectiveness

Optimization criteria

Optimization algorithm(s)

Control philosophy

SENSOR MANAGEMENT TECHNIQUES

is

Sensor Management

does

A process which seeks to manage and coordinate the use of sensing resources in a manner that improves the processof data fusion and ultimately that of Perception synergistically

•Automation of sensor allocation, pointing, moding and emission

•Prioritization and scheduling service requests

•Assistance in sensor data fusion

•Support of reconfiguration and degradation

•Optimization of the use of the available sensor assets

•Assistance in Communication

Based on

1. Heuristics2. Expert Systems 3. Utility Theorem 4. Automated control theory5. cognition6. decision theoretic approaches 7. probability theory 8. stochastic dynamic programming 9. linear programming 10. neural networks 11. genetic algorithms 12. information theory

EXAMPLE OF DESIGN AND EVALUATION OF SENSOR MANAGEMENT SYSTEMS

MOTIVATIONMOTIVATION

ENVIRONMENT PLATFORM CAPABILITY

Dynamic Jamming/Deception Low Observables Numbers/Spacing/Combinatorics

Integrated Avionics Agile/Progammable/Steerable Sensors

limited by

Insufficient sensor resources Varied sensor capability/performance

- Limited processing- Limited power

Unplanned sensor failures Mission constraints (e.g., EMCON)

LEVEL FOUR PROCESSING:PROCESS REFINEMENT

SensorsSensor

Processing

FusionProcessing

Levels

ΣMissionRequirements

MISSION OPERATIONS(e.g. Weapon System) Combat

Environment

EVALUATION

Σ

Fusion Control

Source Requirements

SOURCE AND FUSION CONTROL LOOP

Information Requirements

Info

MISSION MANAGEMENT LOOPMission Effects

MeasurePerformance

and Effectiveness

MOE

MOP

Process Control

DESIGN ELEMENT COUPLING/COMPLEXITY

SENSOR MANAGEMENTSENSOR MANAGEMENTSYSTEMSYSTEM

• Scheduling Parameters• Sensor Synergisms• Areas of Optimum Performance

SENSOR SUITESENSOR SUITE

• Sensor Parametric Agility• Performance Specifications• Contribution to State

SENSOR DATA FUSION SENSOR DATA FUSION PROCESSINGPROCESSING

• Alignment Requirements• Association Options• Tracking Requirements/Options• Computational Aspect

o

o

REQUIREMENTS/PRIORITIESREQUIREMENTS/PRIORITIES

MULTI-SENSOR MANAGEMENT AND UTILITY

SIMILAR TO THE CASE OF AN AGILE SINGLE SENSORSIMILAR TO THE CASE OF AN AGILE SINGLE SENSOR

CONTROL VARIABLES UTILITY GOAL

Time or points on target

Sampling rate

Measured parameters

Search/track trade-offs

Covertness

Maneuvering target characteristics

Target density

High threat target present

Target detection performance

Tracking accuracy

SEARCH UTILITY F[N detected / N undetected]

TRACKING UTILITY CALC COV / DESIGN COV

GENERAL SENSOR CONTROLS

CATEGORY CONTROL FUNCTIONS INPUT TO SENSOR

Mode Control Functions On/Off control Sensor mode selection:

Power level (active sensors) Waveform or processing mode (long-range search, high resolution, NCTR, etc.) Scan, track, or track-while-scan

Sensor processing parameters: Decision thresholds Detection, track, ID criteria

Spatial Control Functions Pointing coordinates (center of field of view (FOV)) Field of view selection Scan/search rate Scan/search pattern select Parameters to control individual sectors:

Sector coordinates Modes within sector

Parameters for designated targets: Target or track index (number) Coordinates or search volume Mode to be used Predicted time of appearance Dwell time on target

Temporal ControlFunctions (timing)

Start/stop times for modes and sector control Specified sensor look time Specified dwell time on target and search Maximum permissible emission duty cycle

Reporting Control Report filters based on target attributes: Friend, foe, or both Filter by class or type Filter by lethality

Report filters based on spatial attributes: Min/max range limits Altitude layer filters Spatial region filters (sectors and volumes defined by geometry)

Priority of designated targets (by index)

FACTORS IN SETTING TARGET PRIORITY

PRIORITY FACTORCATEGORIES

SPECIFIC FACTORS & CRITERIA

IDENTITY Target Allegiance (friend, foe, or neutral) Target type or class Target lethality

INFORMATION NEED Spatial location accuracy Target identification status Track state estimate accuracy Detected targets: – Track filter covariances – Influences revisit rates required to achieve a specified tracking accuracy

– Search volumes, sectors:– Scan or dwell period requirements to achieve a specified detection/ intercept probability

THREAT (DEFENSIVE FACTORRELATIVE TO HOSTILE,UNKNOWN PRESUMED HOSTILETARGETS)

Target type identity Range to target (R) Range rate Range/range rate (time-to-go or imminence) Target lethality (a function of target type) Geometry of own ship relative to target’s weapon envelope(s)

OPPORTUNITY (OFFENSIVEFACTOR RELATIVE TO POSITIVEIDENTIFIED HOSTILE TARGETS)

Geometry of target relative to own (or other friendly) weapon envelopes Range relative to own weapon envelopes Time-to-go until target reaches detection/engagement point Probability of own ship detection by target

FIRE CONTROL NEEDS State of sensor/weapon commitments against target (lock, track, in-flight, etc.) Time-to-go for command-guided weapons in flight

GENERAL SENSOR MANAGEMENT FUNCTIONS

External Controls

EventPrediction

TargetPrioritization• Threat• Opportunity• Information Need

Sensor Prediction

Sensor Performance

Models

Spatial-TemporalCoverage Control

SituationAssessment

Level 2, 3Fusion

Level 1Data Fusion

TargetDataBase

SearchPriorities

Spatial-TemporalParameters

Capacity

Sensor- Target

Assignment

Allocation, Scheduling and ControlTarget

PrioritiesObjectivesFunctions

Control

DataStatus

Sensor Manager Functions

ManualPriorities

EMCon Cues

EMCon Cues

SensorInter face

SensorPerformance

SensorStatus

Walt, E. and Llinas, J., Multisensor Data Fusion”, Artech House, Boston, MA, 1990.

POSSIBLE MOP/MOE LIST FOR EVALUATION OF SENSOR MANAGEMENT

PROCESSDetection and Leakage: Range at target detection Percent of aircraft detected at 100 nmi Percent of aircraft detected at 50 nmi Percent of aircraft detected at 25 nmi Number of undetected targets in the field of regard Percentage of detected targets in the field of regard Time in sensor field of regard prior to detection

Target Tracking: Track variance for hostile aircraft Track variance for priority targets Track variance for all aircraft Number of tracks divided by number of targets Number of track miscorrelations Number of dropped tracks Number of dropped tracks excluding tracking leaving the field of

regard Revisit frequency for priority targets Revisit frequency for hostile targets Revisit frequency for all aircraft Percent of hostile aircraft with launch quality tracks Number of targets tracked divided by the system tracking capacity

Identification: Percent of hostile aircraft correctly identified Range of ID for hostile aircraft Range of ID for all aircraft Time in sensor field of regard prior to identification

Raid Assessment: Percent of targets correctly raid assessed Estimated targets in track/actual targets in track Time in sensor field of regard prior to raid

assessment

Kill Assessment: Percent of targets correctly assessed as killed Percent of targets correctly assessed as alive

Emissions: Total power emitted Power emitted per unit time Number of enemy RWR detections of ownship

Sensor Utilization: Percent of sensor time idle Sensor time spent reacquiring tracks Number of impossible sensor attempts Number of unsuccessful sensor attempts Percent of redundant sensor applications

System Response: Task turn around time Critical request response time Number of starved tasks

Computational Performance: MIPS utilized Memory utilized Bus bandwidth utilized

E-BUSINESS CONCEPTS FOR LEVEL 4

Leveraged Leveraged Information Information TechnologiesTechnologies

• Internet Auctions Internet Auctions & Markets& Markets

• E-CommerceE-Commerce

• Intelligent AgentsIntelligent Agents

• Distributed Distributed Resource Resource AllocationAllocation

E-business concepts are being applied to resource management in multi-sensor systems

MULTIPLE CONSUMER, MULTIPLE SENSOR SCENARIO

processorBattery

Transmission Channel

Bandwidth

Sensor

Manager

Information

Consumers

Information

Suppliers

A ONE TO ONE MAPPING! (ALMOST)

Money

Market

Seller One

Consumer

Consumer

Consumer

Seller Two

Seller Three

AUCTIONS

•Elaborate mechanisms for true consumer preference elicitation

• Sensor network Environment is a co-operative environment

• Appealing mechanism: But other complications exist

WHY STANDARD AUCTION MECHANISMS ARE NOT DIRECTLY

APPLICABLE: CASE 1

Consumer

Consumer

Scan Target at rate r1

Scan Target at rate r2

>r1

Target

Scan a

t r2

Case 1: A single supplier can satisfy multiple consumers with a single product/service

Example: Single item (Scan at rate 2) satisfies both consumers => Each item can be allocated to multiple consumers

Consumer

Consumer

Search Grids 1 to

100

Search Grids 50 to

150

Search

Grids

50 to

100

Case 2: The optimal market solution might require a bid that does not exist.

Example: The sensor capability cannot fully satisfy either consumer scan request.

WHY STANDARD AUCTION MECHANISMS ARE NOT DIRECTLY APPLICABLE: CASE 2

A MARKET-BASED DESIGN

Based on research by T. Mullen et al

Transmissionchannels

Bandwidth

Sensors

Sensor

pointing

Battery power

Processing power

Sellers

Produce

Finished goods Environmental scans, target tracks, .

Consume Consumers

Market Auctioneer

SensorManager

Schedule

Mission Manager

Budgets & Task responsibilities

Mission reports

bid

MARKET MECHANISMS

is

Market Mechanisms

Related to

Design and implementation of resource allocation problems based on some pricing system

Price Equilibrium

Market Conditions whereDemand equals supply

Resource Allocationoptimum

Rely on

is

•Price Tatonement:InherentlyAnytime

•Quantity Tatonement:Anytime adaptationAvailable

calculated

using

REMARKS ON SENSOR MANAGEMENT

Infrastructure technologies will permit integrated, ever-smarter Infrastructure technologies will permit integrated, ever-smarter system architecturessystem architectures This leads to feasibility of sensor management

Sensors are required by more than just the fusion processorSensors are required by more than just the fusion processor This creates contention for sensor service

In an integrated system, coupling exists betweenIn an integrated system, coupling exists between sensor management--fusion--sensors

This leads to a difficult global optimization problem In general, this subject is under-researchedIn general, this subject is under-researched

TOPIC 13 ASSIGNMENTS

Preview the on-line topic 13 materials Read chapter 8 of Hall and McMullen (2004) Read Mullen, et al article (2006)

DATA FUSION TIP OF THE WEEK

Situational Awareness is vitally connected with the effective control of the data fusion process from pointing and controlling the sensors to dynamic selection and control of algorithms to guiding an analyst/user attention