Polymorphous Computing Architectures

Run-time Environment And Design Application for Polymorphous

Technology Verification & Validation (READAPT V&V)

Lockheed Martin Advanced Technology Laboratories1 Federal Street • A&E Building

Camden, New Jersey 08102

Multi-Mission

A

MultipleSensors(A,B,C...X)plug & play

B C X….

Months Days Seconds

Response or Morph Time:

Polymorphous ComputingPolymorphous ComputingArchitecturesArchitectures

Goal: Goal: Computing systems (chips, networks, software) that will Computing systems (chips, networks, software) that will verifiably morph to changing missions, sensor configurations, verifiably morph to changing missions, sensor configurations, hardware failures, and operational constraints during a mission or hardware failures, and operational constraints during a mission or over the life of the platformover the life of the platform

Mission Aware ComputingMission Aware ComputingMission Aware ComputingMission Aware Computing

In-MissionRe-target

-abilityPlatformtransit Multi-sensor

processing

Tracking

• Autonomy is driving Autonomy is driving exponential Growth of exponential Growth of Flight-Safety-Critical Flight-Safety-Critical SystemsSystems

• Mission-critical Mission-critical Applications project Applications project similar growthsimilar growth

Verification & Validation Verification & Validation Problem Problem • V&V currently represents 40 to 70% of total system costsV&V currently represents 40 to 70% of total system costs

• Next generation systems will drive this cost up exponentially Next generation systems will drive this cost up exponentially — Increasing Scale: Multi-mission roles, improved sensors, etc. Increasing Scale: Multi-mission roles, improved sensors, etc. — Increasing ComplexityIncreasing Complexity

– Introduction of autonomy, adaptive mission processing, decision aiding Introduction of autonomy, adaptive mission processing, decision aiding capabilities capabilities

– High percentage of non-deterministic functionalityHigh percentage of non-deterministic functionality– Emergence of adaptive computing (PCA, etc.)Emergence of adaptive computing (PCA, etc.)

• Current V&V solutions are costly with limited scalabilityCurrent V&V solutions are costly with limited scalability

New approaches to V&V are critical to the affordabilityNew approaches to V&V are critical to the affordabilityand adoption of highly adaptive, cognitive systemsand adoption of highly adaptive, cognitive systems

New approaches to V&V are critical to the affordabilityNew approaches to V&V are critical to the affordabilityand adoption of highly adaptive, cognitive systemsand adoption of highly adaptive, cognitive systems

Test (V&V)Test (V&V)

CodeCode

Leveraging Design Models Leveraging Design Models for V&V for V&V

RequirementsRequirements

Production SystemProduction System

Current ApproachCurrent Approach

executable spec.executable spec.

Model Generated SWModel Generated SW

SimulateSimulate

Rapid PrototypeRapid Prototype

ModelModel

model-based V&Vmodel-based V&V

Use the existing Use the existing design models for design models for V&V V&V beforebefore implementation implementation and again during and again during OperationOperation

OperateOperate

Recommended Recommended AdditionsAdditions

High Payoff TechnologiesHigh Payoff Technologies

• Enhanced model-based techniques for complex, dynamic systems Enhanced model-based techniques for complex, dynamic systems — Formal methods: Domain-specific approaches provide provably correct designs; Formal methods: Domain-specific approaches provide provably correct designs;

dramatic (4-10X) reduction in software, integration, and test costsdramatic (4-10X) reduction in software, integration, and test costs— Hybrid methods: Combine discrete (state, flow-based) and continuous (system Hybrid methods: Combine discrete (state, flow-based) and continuous (system

dynamics, temporal) modelsdynamics, temporal) models— Stochastic methods: Complexity- and probability-based characterization of non-Stochastic methods: Complexity- and probability-based characterization of non-

deterministic system functions and softwaredeterministic system functions and software— Evolutionary methods: Mutation-based testing and constraint generationEvolutionary methods: Mutation-based testing and constraint generation

• Constraint-driven specification, implementation, and run-time enforcement of Constraint-driven specification, implementation, and run-time enforcement of emergent behaviorsemergent behaviors

• Information integration across multiple domains/aspectsInformation integration across multiple domains/aspects

Enhanced model-based techniques optimize test coverage with reduced Enhanced model-based techniques optimize test coverage with reduced cost; maximizes V&V payoff for next generation systemscost; maximizes V&V payoff for next generation systems

Enhanced model-based techniques optimize test coverage with reduced Enhanced model-based techniques optimize test coverage with reduced cost; maximizes V&V payoff for next generation systemscost; maximizes V&V payoff for next generation systems

READAPT V&V Thesis and READAPT V&V Thesis and GoalsGoals

• ThesisThesis— Run-Time monitoring and correcting is a revolutionary Run-Time monitoring and correcting is a revolutionary

solution for ensuring “safe” and properly executing solution for ensuring “safe” and properly executing behavior for PCA and cognitive systems-based adaptable behavior for PCA and cognitive systems-based adaptable architecturesarchitectures

• READAPT V&V goalsREADAPT V&V goals— Develop efficient and reliable means of capturing both Develop efficient and reliable means of capturing both

system behaviors and system designs (model behaviors)system behaviors and system designs (model behaviors)— Develop ability to monitor complex, adaptable systems at Develop ability to monitor complex, adaptable systems at

run-time to behaviors captured during the design processrun-time to behaviors captured during the design process— Develop ability to force a complex system into a properly Develop ability to force a complex system into a properly

behaving state in response to changing behaviors and behaving state in response to changing behaviors and behavior violationsbehavior violations

Overall ApproachOverall Approach

• Overall ApproachOverall Approach— Summarize requirements and approaches for avionics Summarize requirements and approaches for avionics

V&VV&V— Investigate additional V&V requirements for PCA Investigate additional V&V requirements for PCA

enabled avionics applicationsenabled avionics applications— Model a PCA architecture for PCA V&V R&DModel a PCA architecture for PCA V&V R&D— Model a non-determinate avionics application Model a non-determinate avionics application

(algorithm) for PCA V&V experimentation(algorithm) for PCA V&V experimentation— Research, develop, and model a hierarchical, integrated, Research, develop, and model a hierarchical, integrated,

reactive-configuration and behavior-monitoring reactive-configuration and behavior-monitoring verification approach for PCA enabled applicationsverification approach for PCA enabled applications

— Integrate models to evaluate the V&V research for a PCA Integrate models to evaluate the V&V research for a PCA enabled avionics applicationenabled avionics application

Application ModelApplication Model

CSIMCSIMCoreCore

PCAPCAPlatformPlatform

ModelModel

Application Application System System

SpecificationSpecification

PCAPCASystem System

SpecificationSpecification

Requirements Requirements SpecificationSpecification

simulationsimulationtracetrace

MEDL MEDL ScriptScript

MaCMaCToolTool

configureconfigureoutputoutput

formalizeformalize

modelmodelmodelmodel

compilecompile

Run-Time Monitoring and Run-Time Monitoring and CorrectingCorrecting

Exhibit DisplaysExhibit Displays

CSIM MaCS

Integrator

Monitoring scriptconfigure

state records interface objects

event def.

Target Set (Endpoint)

Grid Size

Cardinality (Direction)

Planning Horizon

C(r,r+dr)Incremental Cost

Laptop 1 - CSIM/MaCS DisplayLaptop 1 - CSIM/MaCS Display

Route Planning

Algorithm

Laptop 2 -Laptop 2 -Flight with Route Planner Flight with Route Planner

DisplayDisplay

• Flight Simulator provides context displayFlight Simulator provides context display

• CSIM simulates PCA enabled route plannerCSIM simulates PCA enabled route planner

• CSIM controls CDU on Flight SimulatorCSIM controls CDU on Flight Simulator

• MaCS monitors route planner execution & MaCS monitors route planner execution & controls PCA & algorithm adaptationcontrols PCA & algorithm adaptation

Overall Demo ScenarioOverall Demo Scenario

SAM siteDetection & route replan

ApproachingSAM site

Run-time MonitorDetects performanceEnvelope problem;

Revert back toState 2 (Safe)

PassedSAM site

Pre-planned Mission

Execution

Threat Avoidance & Replan

Threat Avoidance &

Targeting Replan

Continuous Replan During Avoidance to

Include Targeting

PCA Morph

State 1:

PCA Morph

State 2:

PCA Morph

State 3:

Threat Avoidance & Replan

PCA Morph

State 2:

PCA Morph

State 4:

States

Events

StreamStream Processors Processors indicated byindicated byfilledfilled boxes boxes

GPGP Processors Processors indicated by indicated by (non-Red) (non-Red) outlinedoutlined boxes boxes

Dynamic bar chart Dynamic bar chart indicating indicating totaltotal active processors, active processors, active active streamstream processors active processors active GPGP processors and processors and active active threadedthreaded processorsprocessors

Total active Total active processor count processor count displaydisplay

MaCS messages MaCS messages and statusand status

Mission statusMission status

PCA Virtual Processor PCA Virtual Processor State and ActivityState and Activity

System State System State and Task Flowand Task Flow

RADAR TasksRADAR Tasks

ImagingImaging

Route Route PlanningPlanning

Self Test Self Test and MaCSand MaCS

Flight Flight ControlControl

Threat Threat AvoidanceAvoidance

Mission Mission AssignmentAssignment

CommsComms

ThreadedThreaded Processors Processors indicated by indicated by REDRED outlinedoutlined boxes boxes

CSIM PCA Flight Control CSIM PCA Flight Control DisplayDisplay