Avionics Compositional System of Systems Simulation and Modeling - - PowerPoint PPT Presentation

avionics compositional system of systems simulation and
SMART_READER_LITE
LIVE PREVIEW

Avionics Compositional System of Systems Simulation and Modeling - - PowerPoint PPT Presentation

Avionics Compositional System of Systems Simulation and Modeling Tool Chain ASSIST October 28, 2019 Tool Expo for Model Based Embedded Systems Development Contact Information: Phillip Suematsu, Dhruv Monga, Howard Warner, Juan Gutierrez


slide-1
SLIDE 1

DISTRIBUTION A. Approved for public release: distribution unlimited.

Avionics Compositional System of Systems Simulation and Modeling Tool Chain ASSIST

October 28, 2019 Tool Expo for Model Based Embedded Systems Development

Contact Information: Phillip Suematsu, Dhruv Monga, Howard Warner, Juan Gutierrez

Physical Optics Corporation 1845 W. 205th Street, Torrance, CA 90501 Phone: 310-320-3088 Email: {psuematsu, dmonga, hwarner, jgutierrez}@poc.com

This work is performed under contract #: W911W6-18-C-0047, W911W6-18-C-0012 W911W6-19-C-0015, W911W6-19-C-0038

DISTRIBUTION A. Approved for public release: distribution unlimited.

slide-2
SLIDE 2

DISTRIBUTION A. Approved for public release: distribution unlimited.

2

  • Founded in 1985
  • Small Business, Employee Owned
  • Financially Strong & Profitable every year
  • 270 employees – 30 Ph.D.s, 112 Engineers
  • Revenue – Over $115M (2019)
  • 117,344 sq. ft. facilities, 4 buildings
  • 2020 Expansion – Additional 53,700 sq.ft., 2 buildings
  • Over 160 issued patents – 60 technologies
  • Strategic Advisory Board

PHYSICAL OPTICS CORPORATION BACKGROUND

slide-3
SLIDE 3

DISTRIBUTION A. Approved for public release: distribution unlimited.

3

DT DTU HD HDVR ML MLS – NA NAS JA JARV RVIS

RF RF/EW

RF RF Filters WI WISDEM PA PALM

Se Sensor/Sc Scanner Cyber Secu curity

X-Ra Ray Da Data Fusion Enc Encryption

Da Data – at at – Re Rest

Enc Encrypted d Data

AI/ AI/Deep Deep Lea Learning

DE DEESTAC OR ORFOM OM DR DREAM AR ARGUS US

Airborne

Da Data Transfer Unit Hi High Def efinition Data & Video eo Rec ecorder er Ne Network Attached Storage Mi Mission Compute ter r – Di Distributed & Reconfigurable Hi High Power er Tunable e Agile e RF Filter ers Wi Wideband Intelligent Spectrum Pr Predistortion Am Amplifier No Non De Destructive Corrosion Inspection Un Unmanned Su Surveillance Or Orbital Fiber Op Optic Production Module Di Digital RF Co Countermeasure

A2 A2D

In Inter erfer eren ence e Cancel eller er

Em Emerging ng Tech chnology

POC AREAS OF FOCUS

slide-4
SLIDE 4

DISTRIBUTION A. Approved for public release: distribution unlimited.

4

PROBLEM STATEMENT AND SOLUTION APPROACH

  • Use of Multicore Processors in Avionics

– Difficult due to inability to verify performance during requirements, design and implementation stages – Analysis of hard real-time and soft real-time requirements needed

  • Solution Approach

– Rigorous specification of requirements and design using Architecture Analysis & Design Language (AADL) – Input data

  • System design and specification in terms of AADL components
  • Avionics system configuration using AADL specifications

– Output

  • Model parameters from AADL specifications
  • Data analysis results
  • Positive match between specifications and designed system
  • Specification violations/contradictions in designed system and deficiencies

– Use of simulation and virtual integration to verify requirements and design

slide-5
SLIDE 5

DISTRIBUTION A. Approved for public release: distribution unlimited.

5

POC SOLUTION

Avionics Compositional System of Systems Simulation and Modeling Tool Chain (ASSIST)

slide-6
SLIDE 6

DISTRIBUTION A. Approved for public release: distribution unlimited.

6

GOAL: ANALYZE DEPENDENCIES AMONG COMPONENTS

Buffer (Overflows) Processor(s) (Execution Time) Threads (Contention, Race) Memory (Availability

  • r Lack)

Bus (Delay, Latency)

Input Output Mutual Dependencies Modeled and Simulated Verify if soft/hard real- time deadlines are met

slide-7
SLIDE 7

DISTRIBUTION A. Approved for public release: distribution unlimited.

7

ASSIST FEATURES

ASSIST Baseline Simulator Components Subcomponents Ports, Flows Bus Access… Unified Error, Behavioral Analysis Indirect Error Propagation Analysis and Identification AADL Behavioral Annex Integrations (EMV2, BA, BLESS, AGREE) Change Impact Analysis Design Ripple Effects in System Architecture What-if Analysis with Objective Optimization

slide-8
SLIDE 8

DISTRIBUTION A. Approved for public release: distribution unlimited.

8

AADL COMPONENT MODELING FUNCTIONALITY

  • AADL Software Components

– Thread, Process: models subprogram execution – Data: models data access latency – Subprogram: statistical model of code execution and data access times

  • AADL Hardware Components

– Processor

  • Scheduler: models thread preemption using priority queue
  • Memory: models context switching and latencies caused by cache misses
  • Device: models sensor and communication components
  • Bus: data exchange mechanism between components
  • AADL Properties

– Timing (Compute execution time, deadline), memory access

  • Component Connections

– Control, data flows – Connection features

  • In/Out/both, direction, ports, (a)synchronous
slide-9
SLIDE 9

DISTRIBUTION A. Approved for public release: distribution unlimited.

9

SUPPORTED AADL FEATURES

  • Package specification

– Annex libraries not processed

  • Import declaration
  • Component Types

– Software Category: Subprogram, Thread, Process – Execution Category: Memory, Processor, Bus, Device – Composite Category: System – Features – Flows – Properties – Extends

  • Component Implementations

– Subcomponents – Calls – Connections – Flows – Properties

  • Subcomponents

– Array dimensions – Refined to

  • Port support only
  • Features

– Direction: in, out, in out – Ports: event, data, event data – Requires – Provides

  • Subprograms

– Call sequence – Execution time

  • Bus access connections
  • Flow specifications

– Types: source, sink, path

  • End-to-End Flow specifications
  • Property Sets
  • Property Types

– Basic data types, Reference, Record

slide-10
SLIDE 10

DISTRIBUTION A. Approved for public release: distribution unlimited.

10

SOLUTION

  • Avionics Compositional System of Systems Simulation and

Modeling Tool Chain (ASSIST)

  • Analysis of hard real-time and soft real-time requirements

– Aviation system of systems simulation using representative use case – Generating configuration for simulation – Verification of system against architecture model defined in AADL

  • Approach

– Discrete event simulation of an SoS with multi-core processors – Input data: AADL specifications, external data sources – Output

  • SoS model characterized by parameters from AADL specifications
  • Data analysis results
  • Positive matches between specifications and designed system features
  • Specification violations/contradictions in designed system and

deficiencies

slide-11
SLIDE 11

DISTRIBUTION A. Approved for public release: distribution unlimited.

11

ASSIST HIGH LEVEL ARCHITECTURE

slide-12
SLIDE 12

DISTRIBUTION A. Approved for public release: distribution unlimited.

12

MULTICORE PROCESSORS IN AVIONICS

  • ASSIST design motivated by Multicore Processor Analysis

– Federal Aviation Administration Study - Assurance of Multicore Processors in Airborne Systems

http://www.tc.faa.gov/its/worldpac/techrpt/tc16-51.pdf

  • Statistics recommended by FAA and collected by ASSIST:

– Core utilization (% utilized averaged over ms) – Processing time per sensor message – Processing time per thread – Cache miss (+hit) counts and miss (+hit) rates/ms – Thread execution details:

  • Assigned processor
  • State transitions (running, executing, waiting on resource, idle)

– Deadline violations – Flow rates per message

slide-13
SLIDE 13

DISTRIBUTION A. Approved for public release: distribution unlimited.

13

DEMONSTRATION SCENARIO (AADL MEETING OCT 2019, WASHINGTON DC)

  • Virtual Integration via ASSIST Simulation featuring Hardware In

the Loop (HWIL)

  • Hardware: POC’s flight data recorder

– Current input sensors: turbine, fire, acceleration, altitude – (Modified) system design includes:

  • Multi-Core CPU, RAM, Caches, Bus
  • An additional video sensor (live feed)
  • Software: Data processing framework

– Threads, processes, subprograms to record data from sensors

  • Scenario #1:

– Flight Data Recorder (FDR) simulation using a dual core system

  • Simulate feeds from sensors (timing, message-size, order of message arrival

modeled )

– Additional messages from video: data-size, timing

  • Scenario #2

– Perform joint FDR+video simulation using a quad core system

slide-14
SLIDE 14

DISTRIBUTION A. Approved for public release: distribution unlimited.

14

CHALLENGES IN SYSTEM IMPLEMENTATION

  • Mismatch in simulation rate and data-arrival rate

– Require tradeoff between simulation times and modeling fidelity

  • Running multiple simulations simultaneously not

possible on a dedicated laptop

  • Modeling large systems will require platforms with

high computational capabilities

  • Ease of software distribution among stakeholders for

evaluation during Capstone event

  • Computing Approach Scalable with Size and

Complexity of Simulations Is Needed

slide-15
SLIDE 15

DISTRIBUTION A. Approved for public release: distribution unlimited.

15

DEMONSTRATION SETUP

Flight Data Recorder Turbine Sensor Fire Sensor Acceleration Sensor Video Camera Altitude Sensor 20+ sensors ASSIST Simulator Results Display

TCP/IP Network Software APIs

Time series Database

slide-16
SLIDE 16

DISTRIBUTION A. Approved for public release: distribution unlimited.

16

DEMONSTRATION RESULTS

  • Dual Core:

– Without video camera: no violations, normal/expected

  • perating behavior

– Adding a video camera:

  • CPUs are unable to process additional data.
  • Limited computation capabilities result in deadline violations and

increased processing time for all critical sensor tasks.

  • Quad Core:

– Without video camera: no violations, normal/expected

  • perating behavior

– Adding a video camera:

  • Data rate is still too high for any single CPU to handle
  • However, additional cores are available so the critical sensor

processing tasks are not affected

slide-17
SLIDE 17

DISTRIBUTION A. Approved for public release: distribution unlimited.

17

RESULTS – DUAL CORE

slide-18
SLIDE 18

DISTRIBUTION A. Approved for public release: distribution unlimited.

18

RESULTS – QUAD CORE

slide-19
SLIDE 19

DISTRIBUTION A. Approved for public release: distribution unlimited.

19

COMPARATIVE ANALYSIS

Dual Core Quad Core

Processor Core Utilization

  • Dual Core: 72% / 64%
  • Quad Core: 45% / 44% / 78% /

47% (additional headroom) ✓ Thread Preemption Latency

  • Dual Core: ~15-20 instances

preemption exceeds 10%

  • Quad Core: 0 instances

preemption exceeds 10% ✓ Processing Deadline Violations

  • Dual Core: 4 critical threads

100’s violations

  • Quad Core: only non-critical

video thread ✓

slide-20
SLIDE 20

DISTRIBUTION A. Approved for public release: distribution unlimited.

20

NEXT STEPS

  • ARINC 653

– Virtual Processor Partitioning

  • Additional Summary Statistics
  • AADL Features

– Parameters – Access to peripherals – Programming languages

  • GUI Improvements

– Improve interface to add files and simulate. Most of the technical work is done – just need to clean it up. – Improve GUI for comparing and contrasting variations in SoS AADL models

  • Cloud Infrastructure – deploy as cluster on Amazon

– Currently working on deploying on Kubernetes