Upload
others
View
8
Download
0
Embed Size (px)
Citation preview
© 2019 The MathWorks, Inc. 1
A look to the future with
Model-Based Design
Andy GraceVice President of Engineering
Design Automation
2
MathWorks Today
in 2018 revenues with
60% from outside the US
$1B+4500+
staffin 31 offices around
the world
3 million+
usersin more than 180
countries
and profitable every year
Privately
held
Headquarters
Natick, MA USA Europe
France
Germany
Ireland
Italy
Netherlands
Spain
Sweden
Switzerland
UK
Asia-Pacific
Australia
China
India
Japan
Korea
North America
United States
© 2019 The MathWorks, Inc. 3
Technology Megatrends Driving Automotive
1. Vehicle Electrification
2. Autonomous Driving
3. Connected Vehicles
Software everywhere
© 2019 The MathWorks, Inc. 4
Software is reshaping the automotive industry
Marc Andreessen
Founder of Netscape,
Renowned Venture capitalist
In the future every
company will become a
software company
© 2019 The MathWorks, Inc. 5
Augmenting control with
machine learning (BMW)
Trailer backup assist (Ford) Autonomous driving (Voyage)
Software is reshaping the automotive industry
© 2019 The MathWorks, Inc. 6
Agile Values
Individuals & Interactions over Process and Tools
Working Software over Comprehensive Documentation
Customer Collaboration over Contract Negotiation
Responding to Change over Following a Plan
- The Agile Alliance, 2001
“While there is value in the items on the right, we value the items on the left more.”
© 2019 The MathWorks, Inc. 7
Agile: Values, Principles and Practices
4 VALUES 12 PRINCIPLES PRACTICES
Agile is a mindset defined by values, guided by principles and manifested through many different practices.
Agile practitioners select practices based on their needs.
~ Agile Practice Guide (PMI® and Agile Alliance®)
© 2019 The MathWorks, Inc. 8
Typical agile development workflow
1-4
Weeks
© 2019 The MathWorks, Inc. 9
Models Understanding==
© 2019 The MathWorks, Inc. 10
© 2019 The MathWorks, Inc. 11
Simulation Physical Prototyping
© 2019 The MathWorks, Inc. 12
Simulation is key to Level 4-5 autonomy
Knowns Unknowns
Critical situations are in the long-tail*Simulation helps achieve this improbable task
*Source: Center for Artificial Intelligence, Saarland University
© 2019 The MathWorks, Inc. 13
Simulation
Coding Verification
Modeling
Automation
Model-Based Design Systematic use of models throughout the development process
Fast repeatable tests
Fast agile
development loops
© 2019 The MathWorks, Inc. 14
Types of models
Components
Software Physics
Systems
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 15
Physical components
Communications Channel
Vehicle Component Sensor Model
Motor
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 16
Simscape for physical modeling
Publication-quality
diagramsModels just runSimscape modeling
language
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 17
Types of models
Components
Software Physics
Systems
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 18
Simulink as an Integration Platform
MODELING & SIMULATION
Simulink
© 2019 The MathWorks, Inc. 19
Simulink as an Integration Platform
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 20
MODELING & SIMULATION
Simulation Integration: Infrastructure
VisualizationMulti-actor
Scenarios
Vehicle
Configuration
Solver
Technology
Data
Management
© 2019 The MathWorks, Inc. 21
MODELING & SIMULATION
Simulation Integration: Analyses
© 2019 The MathWorks, Inc. 22
MODELING & SIMULATION
Parallel simulations Simulation ManagerX 1,000,000’s
Scaling up simulations
Programmatic test
creation
© 2019 The MathWorks, Inc. 23
Types of models
Components
Software Physics
Systems
“A typical ECU contains 2000
function components that each
are developed by a different
person.”
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 24
Working at a high-level of abstraction
SOFTWARE COMPONENTS
MATLAB Simulink Stateflow
© 2019 The MathWorks, Inc. 25
Component modeling
Startup and shutdown behavior Variant managementReusable
components that
can be adapted to
any software
system
SOFTWARE COMPONENTS
…
……Reusable
Component
…
Reusable
Component
© 2019 The MathWorks, Inc. 26
Types of models
Components
Software Physics
Systems
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 27
System architecture is the #1 topic
Breakout Topic Requests (2018)
51
55
56
64
75
Efficiency of Generated Code
Testing Simulink Models
Customizing Embedded Coder
Sensor Fusion and Tracking
Modeling System Architecture
103
106
123
167
173
Improved UI
Verification & Validation
Large-scale Modeling
Code Generation
System Architecture
SOFTWARE & SYSTEM ARCHITECTURE
Feature Prioritization (2017)
© 2019 The MathWorks, Inc. 28
Systems
Systems engineering
Requirements Components
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 29
Requirements Components
System Composer
Systems engineering
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 30
System
REQUIREMENTS
Component
DESIGNComponent
VERIFICATION
System
ARCHITECTURE
System
SIMULATION
Component
IMPLEMENTATION
System TEST
Linking top-down and bottom-up workflows
SOFTWARE & SYSTEM ARCHITECTURE
© 2019 The MathWorks, Inc. 31
Types of models
Components
Software Physics
Systems
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 32
MODELING & SIMULATION
Controls WirelessSignal Processing RoboticsVision
Deep solutions
© 2019 The MathWorks, Inc. 33
Automotive Products
MODELING & SIMULATION
Deep solutions
Powertrain Vehicle
CalibrationAutomated Driving
© 2019 The MathWorks, Inc. 34
MODELING & SIMULATION
Automotive Reference Applications
Lane Keeping AssistPure EV
Car Vehicle DynamicsHybrid Powertrain
© 2019 The MathWorks, Inc. 35
Deep solutions for autonomous systems
Perception
Localization Planning
Control
Semantic Segmentation (17b)
Automated Driving
System Toolbox
Path Planning (19a)
Automated Driving ToolboxSLAM (18a)
Robotics System Toolbox
Adaptive Cruise Control (17a)
Automated Driving
System Toolbox
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 36
Lane Keep Assist
Model Predictive Control
Automatic Emergency Braking
Automated Driving Toolbox
Deep solutions for autonomous systems
© 2019 The MathWorks, Inc. 37
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 38
DeployTrain
MATLAB Workflow for Deep Learning:
FROM SCRATCH
TRANSFER
Access Data Preprocess Access Models
MUNGING/LABELING
FUSION
DENOISING
BUILD
BORROW
Deep Learning ToolboxCreate, analyze, and train deep learning networks
Interoperability with open source
networks
Deep Network Designer App Inference performance
Network training performance Deployment supportDomain-specific workflow
support
Ground truth labeling apps for:
• Video
• Audio
• application-specific datastores
© 2019 The MathWorks, Inc. 39
▪ Application examples
Artificial Intelligence for your applications
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 40
▪ Application examples
▪ Control design
Reinforcement Learning Toolbox
Artificial Intelligence for your applications
MODELING & SIMULATION
© 2019 The MathWorks, Inc. 41
Simulation
Coding
Modeling
Automation
© 2019 The MathWorks, Inc. 42
Coding
© 2019 The MathWorks, Inc. 43
Solutions for Vision and Deep Learning
CODE GENERATION
GPU
Fastest
FPGA / ASIC
Lowest Power
CPU
Low Cost
© 2019 The MathWorks, Inc. 44
Model-Based Design C/C++
vs
▪High level of abstraction
▪Advanced analysis tools
▪Automatic code generation
Hand Code
CODE GENERATION
© 2019 The MathWorks, Inc. 45
vs
Model-Based Design C/C++ Libraries
Vendor Libraries
▪No wrappers
▪No data typing
▪No data copies
Hand Code Internal Libraries
CODE GENERATION
© 2019 The MathWorks, Inc. 46
Model-Based Design C/C++ Libraries
Middleware
Hand Code Vendor LibrariesInternal Libraries
▪No wrappers
▪No data typing
▪No data copies
CODE GENERATION
© 2019 The MathWorks, Inc. 47
Simulation
Coding
Modeling
Automation
Verification
© 2019 The MathWorks, Inc. 48
TEST & VERIFICATION
Automated Test and Verification
Find bugs Manage tests Check & Coverage Inspect code
Simulink Design Verifier
Polyspace Bug FinderSimulink Test Simulink Check
Simuink Coverage
Simulink Code Inspector
© 2019 The MathWorks, Inc. 49
Online Access for Test and Verification
TEST & VERIFICATION
Polyspace
Server
Polyspace AccessCONTINUOUS
INTEGRATION
BUG
TRACKINGAUTHENTICATION
DATA
STORAGE
Web browser
© 2019 The MathWorks, Inc. 50
Simulation
Coding Verification
Modeling
Automation
Fast repeatable tests
Fast agile
development loops
Model-Based Design Systematic use of models throughout the development process
© 2019 The MathWorks, Inc. 51
Mechanical-centric Model-centric Software-centric
Who will be successful in the future?
Comprehensive models
Simulation based testing
Generate code and automate verification
© 2019 The MathWorks, Inc. 52
Enjoy the conference