Sameer Prabhu, MathWorks
As the adoption of embedded control systems has increased, software for applying Model-Based Design has also matured in sophistication. Earlier tools were focused on developing innovative controls features. Now, the focus is on two axes: developing innovative features and ensuring the tools are a key part of the overall vehicle systems development process, which encompasses technical, process, and organizational aspects. This presentation starts with a historical overview of automotive trends and Model-Based Design and illustrates how both have evolved over the years. The presentation then looks at the key challenges today in developing innovative automotive technologies and what MathWorks is doing to support deployment of Model-Based Design. It concludes with a look to the future evolution of automotive systems development.
Bill Ortner, Autoliv Corporation
Automotive manufacturers depend more and more on their tier 1 and tier 2 automotive suppliers for electronic products. The automotive electronics market, with its dependence on software engineering skills, has exploded over the past 25 years. This trend has uncovered weaknesses in product development—poor product quality, lack of product standards, and missed customer delivery. The genesis of AUTOSAR, primarily led by European automotive manufacturers, is aimed at addressing and improving automotive electronics and software development through product standardization and reuse.
Autoliv has delivered a number of automotive safety electronic modules to European automotive manufacturers using various version of AUTOSAR. AUTOSAR product architecture with Model-Based Development using auto code generation addresses weaknesses in:
- Software process rigor
- Complex automotive electronig architecture management
- Software reuse
- Software interchangeability
- Software modularity
- Software development methods
Vinod Reddy, MathWorks
A process assessment is a catalyst for process improvement and is often used by organizations to identify what changes to implement and how to phase in the changes. However, it is challenging to assess a model-based software development process due to lack of an established framework. To address this issue, MathWorks has developed such a framework and successfully applied it to organizations worldwide to improve their processes. This presentation will provide an overview of the framework.
Nathan Fitzgerald, John Deere
For John Deere Electronic Engine Controls, performing virtual validation and verification was a key driver in the adoption of model‑based software development. Recent tool improvements provide an understanding of the model test coverage and provide the ability to generate test cases to improve that coverage. While understanding the test coverage of the design can be informative, a higher level goal is to understand test coverage of the requirements of the component requirements. This presentation discusses the use of a reference model along with Simulink Design Verifier™ to generate test cases for the requirements.
This approach has several fundamental benefits. The test cases are based on the requirements and not the implementation, so the design model test coverage metrics provide insight into the completeness of the design. Since the reference model computes expected outputs, the test cases are executed against both the reference model and the design model. The test cases can be executed with an initial design model to find design issues and again with an elaborated model to find implementation issues prior to code generation. Finally, the reference model can be used with any set of inputs to enhance understanding of system operation and the requirements.
Arvind Jayaraman, MathWorks
After years of development, AUTOSAR is now in production. This presentation will provide the latest status on the AUTOSAR standard, a summary of the AUTOSAR tool chain from MathWorks including the AUTOSAR Target Production Package, and case studies from European automotive OEMs and suppliers.
Gary Ferries, General Motors
The freely available document “Control Algorithm Modeling Guidelines Using MATLAB®, Simulink®, and Stateflow®,” has become the de facto standard for companies doing control algorithm modeling. Last year, the North American MathWorks Automotive Advisory Board (NA-MAAB) Style Guidelines Working Group issued a minor revision to the document (version 2.2). The group is currently completing a major revision that is planned for publication later this year (version 3.0). This session presents a brief history of the document and a high level overview of the recent changes.
Danaan Metge, BPG Werks
This presentation includes a description of the history of BPG Werks and its founder, Benjamin Gulak. The current and past versions of the UNO, the philosophy behind the current model, and the future of the machine are described. The session also addresses the technical challenges that came up during development of the UNO and describes how MATLAB and Simulink helped to solve these challenges. The presentation includes a brief case study of the UNO's the tilt system and a description of how BPG used Simulink and SimMechanics™ to improve this system and bring a dynamic, innovative prototype to life.
Tony Mansour, Lear Corporation
This presentation covers some of the practical experience Lear has gained using Polyspace products to improve the quality of their body controller software. Polyspace products are powerful engineering tools that can assist in uncovering hidden defects in products with embedded software. This talk will cover Lear's experience in dealing with the learning curve associated with this analysis tool. We discuss the kinds of problems that are best addressed by this tool and the trade-offs required to get the most benefit. The presentation will also touch on ideas for integrating Polyspace tools into an existing engineering development process.
Collaborating with Model-Based Design: Leveraging Executable Specifications for Body Electronics Applications
Shawn Kalinowski, MathWorks
Executable specification is a well-adopted concept and a key benefit provided by Model-Based Design. This presentation provides an overview of the latest tool capabilities that would make executable specifications a more effective part the OEM/supplier interface, support design sign offs, and generate design artifacts. The capabilities covered include requirement management interface, data management, Simulink Projects, connection to an HMI, and report generator.
A fast and accurate motor model is essential for developing motor controls using Model-Based Design. This presentation describes a process through which such a control-oriented model is developed and verified against experimental data through a series of four physical tests. Specific topics covered include the physical effects that must be captured in the model versus those that can be neglected and test procedures used to identify the value for each of the critical model parameters.
Greg Wolff, MathWorks
This master class uses a field-oriented control algorithm for a permanent magnet synchronous motor to illustrate how to generate efficient C code from the controller model, integrate it with hand code for the embedded device drivers, and use the fully integrated software to spin the motor hardware. Topics include model architecture, algorithm export, scheduling techniques, code profiling, and code verification using processor-in-the-loop (PIL) testing.
Grant Martin, MathWorks
Computer vision uses images and video to detect, classify, and track objects or events in order to understand a real-world scene. Through technical demonstrations, we will show how the MATLAB product family can be used to accelerate algorithm development and exploration and reduce development time for computer vision applications. Topics covered in this master class include:
- Working with files and live sources
- Blob/point detection, feature extraction, and matching techniques
- Video motion analysis with optical flow, and block matching
- Video stabilization and stereo image rectification
- Classification algorithms to recognize image content
- Video display and graphic overlay
- Multicore PC and NVIDIA GPU simulation and acceleration
- Integration with OpenCV