Acumen A Cyber-Physical (CPS) Modeling Language Rigorous - - PowerPoint PPT Presentation

▶
acumen a cyber physical cps modeling language rigorous
SMART_READER_LITE
LIVE PREVIEW

Acumen A Cyber-Physical (CPS) Modeling Language Rigorous - - PowerPoint PPT Presentation

Oct 29, 2022 433 likes •900 views

Acumen A Cyber-Physical (CPS) Modeling Language Rigorous Simulation Walid Taha Halmstad University and Rice University Rigorous Simulation Aaron Ames, Kevin Atkinson , Jerker Bengtsson, Raktim Bhattacharya, Paul Brauner, Robert Cartwright,

slide-1
SLIDE 1

Acumen A Cyber-Physical (CPS) Modeling Language

Walid Taha Halmstad University and Rice University

Rigorous Simulation

slide-2
SLIDE 2

Aaron Ames, Kevin Atkinson, Jerker Bengtsson, Raktim Bhattacharya, Paul Brauner, Robert Cartwright, Alexandre Chapoutot, Adam Duracz, Jan Duracz, Henrik Eriksson, Veronica Gaspes, Christian Grante, Jun Inoue, Michal Konecny, Marcie O’Malley, Travis Martin,Jawad Masood, Marisa Peralta, Cherif Salama, Walid Taha, Edwin Westbrook, Fei Xu, Yingfu Zeng, Yun Zhu Halmstad University, Rice University and Texas A&M, SP, Volvo Trucks

Rigorous Simulation

slide-3
SLIDE 3

Robot Design

slide-4
SLIDE 4

NG-Test Project

slide-5
SLIDE 5

Simulation in innovation

Idea Product Prototype Model

slide-6
SLIDE 6

Simulation in innovation

Idea Product Prototype Model Bug Flaw Disaster

slide-7
SLIDE 7

Rigorous Simulation

Idea Product Prototype Model Bug Flaw Disaster Late “debugging” can be extremely expensive Debugging models (simulation) is a productivity bottleneck

Simulate Test

CPS models must combine C & P!

slide-8
SLIDE 8

This talk

  • The CPS simulation domain
  • Available tools
  • Rigorous simulation
  • The staging connection
  • The E-L equation (previous work)
  • Binding time analysis (ongoing)
slide-9
SLIDE 9

Modeling Simulation

slide-10
SLIDE 10

Why simulation is hard

  • Solving continuous equations:
  • ODEs, DAEs, PDEs, IDEs, ...
  • IVP vs. BVP
  • Hybrid aspects pervasive (both C & Ph)
  • Dealing with precision in models
  • Uncertainty (intervals)
slide-11
SLIDE 11

Hybrid systems basics

Derivatives Computability Discontinuities Interaction

slide-12
SLIDE 12

Specific problems

  • Equality & zero crossing (if x=0 then ...)
  • Zeno-behavior (bouncing ball)
  • Static verification (e.g. solvable, stable)
  • Numerical precision and validity
  • Pole detection (e.g. x’=x*x, x(0)=1(?))
  • Semantic treatment desperately needed
slide-13
SLIDE 13

Current tools

  • Simulink
  • Mathematica, Maple
  • ML, Haskell
  • HA, e.g. CHARON
  • Modelica, Verilog-AMS, ...
slide-14
SLIDE 14

Strengths

  • Simulink: Lots of models (IP)
  • Mathematica, Maple: Symbolics
  • ML, Haskell: Expressivity, concurrency
  • HA, e.g. CHARON: Formal proof
  • Modelica, Verilog-AMS: Equations
slide-15
SLIDE 15

Weaknesses

  • Simulink: Numeric semantics...
  • Mathematica: Executability
  • ML, Haskell: “Indeterminism”
  • HA, e.g. CHARON: Scalability
  • Modelica, Verilog-AMS: Semantics...
slide-16
SLIDE 16

Acumen

  • Acumen ’09: continuous language
  • “Math as a programming language”
  • Acumen ’10 (- now): hybrid language
  • Hierarchical hybrid systems
  • IDE (automatic plotting & 3D view)
  • Rigorous semantics
slide-17
SLIDE 17

Basic Look and Feel

slide-18
SLIDE 18

Hybrid Dynamics Example

slide-19
SLIDE 19

Example for CPS course

slide-20
SLIDE 20

Rigorous Semantics

x E t t t x’

slide-21
SLIDE 21

Rigorous Semantics

slide-22
SLIDE 22
slide-23
SLIDE 23

The Staging Connection

A passive robot walks naturally down inclines Can this be generalized?

slide-24
SLIDE 24

The Staging Connection

  • Effort by one of our three domain experts
  • Robot model uses 8x8 Lagrangian
  • Must convert to “executable” math
  • Mathematica gave a 13MB derivative!
  • It’s very hard for robotics experts to

find the right tools for simulation

slide-25
SLIDE 25

Equational models

slide-26
SLIDE 26

Equational models

slide-27
SLIDE 27

Equational models

slide-28
SLIDE 28

Equational models

slide-29
SLIDE 29

Equational models

slide-30
SLIDE 30

Equational models

slide-31
SLIDE 31

Equational models

slide-32
SLIDE 32

From Acumen to DAEs

slide-33
SLIDE 33

From Acumen to DAEs

slide-34
SLIDE 34

Effect on Performance

Acumen ’09 [ICCPS’10]

slide-35
SLIDE 35

BTA: Syntax

slide-36
SLIDE 36

BTA: Collecting constraints

slide-37
SLIDE 37

BTA: Solving constraints

slide-38
SLIDE 38

Example: Pendulum

slide-39
SLIDE 39

Example: Pendulum

slide-40
SLIDE 40

Example: Pendulum

slide-41
SLIDE 41

BTA: Next Steps

  • Formalizing correctness criteria
  • “Well-annotated programs don’t go

wrong”

  • Establishing sufficiency for an

interesting class of models

  • Example: Rigid body dynamics
slide-42
SLIDE 42

Conclusions (1/2)

  • Rigorous simulation is a powerful tool
  • Being based on simulation makes intuitive
  • Being rigorous makes it a verification tool
  • Semantics makes the tool rigorous
  • Staging implements the semantics efficiently
slide-43
SLIDE 43

Conclusions (2/2)

  • Rigorous simulation naturally accommodates

parametric uncertainty

  • Modeling uncertainty makes simulations

much more informative

  • Using rigorous simulation during early-stage

design has a distinctive flavor that promotes robust design

slide-44
SLIDE 44

Thank you!

  • To download our papers:
  • http://effective-modeling.org
  • To download Acumen
  • http://acumen-language.org
  • For CPS Lecture Notes
  • http://bit.ly/LNCPS-2014