International Automotive Conference 2006International Automotive Conference 2006May 16, 2006May 16, 2006

EMMESKAY, INC.Advanced Technology Solutions

Converting Legacy Converting Legacy Embedded Control Software Embedded Control Software to Executable Specifications to Executable Specifications

Koichi UedaKoichi UedaTOYOTA MOTOR CORPORATIONTOYOTA MOTOR CORPORATION

Yoshitaka UematsuYoshitaka UematsuDENSO CORPORATIONDENSO CORPORATION

Michael BalohMichael BalohEMMESKAY INC.EMMESKAY INC.

2

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

3

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

4

BackgroundBackground

Power Train Requirements are becoming increasingly strict.- Fuel Economy- Clean Exhaust Gas Emission- Performance (e.x. Drivability)

Increasing complexity of Engine Control AlgorithmsIncreasing complexity of Engine Control Algorithms

Development process improvement is an urgent issue.

Keywords:Keywords: ModelModel--Based Development (MBD)Based Development (MBD)

5

MBD ConceptMBD Concept

Real World

Control SoftwareControl SoftwareSpecificationSpecification==

Engine PerformanceEngine PerformanceSpecificationSpecification==

Validation

Validation

Combination

Combination

Engine ModelEngine Model Controller ModelController Model

Engine(Engine, Actuators, Sensors)

Controller(Hardware, Software)

HILSHILSRapid Prot. ECURapid Prot. ECU

SILSSILS

Virtual World

6

MBD ProcessMBD Process

Validation

Control SoftwareSpecification C-code Legacy

C-code

Control design Integration

System Specification Controller

System design Combination

Requirements &Constraints

Control System(Engine + ECU)

Verification

Verification

Automatic Code Generation

7

Major Activities in MBDMajor Activities in MBD

- Model-Based Control- Model-Based Calibration- Model-Based Verification & Validation- Rapid Modeling- Executable Specification- Automatic Code Generation (ACG)

TOYOTA and DENSO have already deployed an ACG environment using Real-TimeWorkshop Embedded Coder

®for advanced and mass production development.

(IAC 2004, June 15, 2004)

Focus:Focus: Executable SpecificationExecutable Specification

8

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

9

Executable SpecificationExecutable Specification

20 20 to 30 % of development period have been reduced.to 30 % of development period have been reduced.(Executable Specification + ACG)

Executable Specification+ ACG

+ ACG+ Hand written codeExecutable Specification

2000 2002 2004 2006

Advanceddevelopment

Productiondevelopment

MBD Application to power train control ECU softwareMBD Application to power train control ECU software

Deployment of SimulinkDeployment of Simulink®®

modelsmodelsas production specificationsas production specifications

NOTE: Executable specifications are applied to roughly 10 % ofNOTE: Executable specifications are applied to roughly 10 % ofengine control algorithms.engine control algorithms.

10

MotivationMotivation

In order to shift to MBD entirely(i.e., further improvements in productivityfurther improvements in productivity),we started a project to convert documentedlegacy control algorithms into executable specifications.

Documented legacycontrol algorithms Executable specifications

11

ExpectationExpectation

Converted Controller ModelsConverted Controller Models

DevelopmentDevelopmentin the virtual worldin the virtual world- Control algorithms can

be explored usingcontroller modelsand plant models.

Other ActivitiesOther Activities- Controller models

enablere-architecting thecontrol modulessoftware efficiently.

MaintenanceMaintenance- MBD process can be

adopted for themaintenance of legacycontrol algorithms.

12

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

13

Key ConsiderationsKey Considerations

- Correct models- Uniform quality- Readable models- Uniform appearance of models- Efficient conversion methodology

-- Structured ProcessStructured Process-- AutomationAutomation

Key to success:Key to success:

14

Basic Idea of ConversionBasic Idea of Conversion

Documents

“Properties for Parameters”“Compiler Configurations”“Style Guidelines”“Conversion Rules” etc.

Additional Information

Conversion SimulinkModel

“Legacy Embedded Software”

C-code

“Executable Specification”

15

Conversion ProcessConversion Process

Architecture Extraction

Product Release

Update Knowledge Base

Model Implementation Test Vector Generation

Verification & Peer Review

Direct PathConditional

Start C-code

SimulinkModel

16

Key Features of ProcessKey Features of Process

- Structured process- Parallel works- Automation throughout the process- Continuous improvement (“Kaizen”)

Similar to ““Assembly LineAssembly Line””

This process is very efficient while delivering good quality.This process is very efficient while delivering good quality.

17

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

18

Model ImplementationModel Implementation

Architecture Extraction

Product Release

Model ImplementationModel Implementation Test Vector Generation

Verification & Peer Review

Start C-code

SimulinkModel

Architecture ExtractionArchitecture Extraction

Product ReleaseProduct Release

Model ImplementationModel Implementation Test Vector Generation

Verification & Peer ReviewVerification & Peer Review

Start C-codeStart C-code

SimulinkModel

SimulinkModel

Process rules and Trained engineers

- Style guidelines- Style checkers

- Customblock sets

- Automationscripts

Target modelsTarget modelsand modelingand modeling

-- CorrectCorrect-- ReadableReadable-- UniformUniform-- EfficientEfficient

19

Quality ControlQuality Control

Architecture Extraction

Product Release

Model Implementation Test Vector GenerationTest Vector Generation

Verification & Peer ReviewVerification & Peer Review

Start C-code

SimulinkModel

Architecture ExtractionArchitecture Extraction

Product ReleaseProduct Release

Model Implementation Test Vector GenerationTest Vector Generation

Verification & Peer ReviewVerification & Peer Review

Start C-codeStart C-code

SimulinkModel

SimulinkModel

Process rules and Trained engineers

- Verificationtechnologies

“numerical correctness”

- Peer review

“functional accuracy”“adherence to style guidelines”

Guarantee ofGuarantee ofthe model qualitythe model quality

-- CorrectCorrect-- ReadableReadable-- UniformUniform-- EfficientEfficient

20

Test Vector Generation (TVG)Test Vector Generation (TVG)

Architecture Extraction

Product Release

Model Implementation Test Vector GenerationTest Vector Generation

Verification & Peer Review

Start C-code

SimulinkModel

Architecture ExtractionArchitecture Extraction

Product ReleaseProduct Release

Model Implementation Test Vector GenerationTest Vector Generation

Verification & Peer Review

Start C-codeStart C-code

SimulinkModel

SimulinkModel

Meaningful test vectors must be found efficiently.Meaningful test vectors must be found efficiently.

- TVG is constructed fromautomatic data generationand manual activity.

- Automatic generation ofstatic signal step tests isused.

- Original coverage metricsare applied.

- TVG continues untilcoverage requirements aremet.

Start

Automatic Test Data Generation

ManualActivity

Coverage OK

no

yes

End

21

ContentsContents

1. Introduction

2. Motivation

3. Conversion Process

4. Fundamental Process Activities

5. Current Plan and Future Works

6. Conclusions

22

Current PlanCurrent Plan

Engi

ne c

ontr

ol

algo

rithm

spe

cific

atio

ns

Current

with other activities in MBDwith other activities in MBD

After several years

Documented spec.

Model implemented spec.

0

100%

Aim:Aim: Two times increase of productivityTwo times increase of productivity

23

Future WorksFuture Works

- Continuous improvement of the conversion processefficient and practical verification tools environment

- Large scale modelingpractical environment, style guidelines

- Integrated development environmenttool chain enhancement

24

ConclusionsConclusions

- Due to further improvements in productivity, TOYOTAstarted a project to convert documented legacyalgorithms into executable specifications.

- Adherence to a structured process is one of the keyenablers for the successful execution of this project.

- The result of this project will bring about greatimprovements to the power train control systemdevelopment process in TOYOTA.

25

Thank you for your attention!Thank you for your attention!