Automotive

MathWorks International Automotive Conference 2006 - Speakers & Presentations

Keynote

Managing the Complexity of E/E Architectures

Dr. Thomas Scharnhorst – Volkswagen

Director of E/E Architectures and Concepts
and Managing Director of Carmeq GmbH, Berlin, Germany

Customer Presentations

Model-Based Motorsports Engine Control Unit
Wolfgang Aßfalg, Alexander Leuze, Bosch – Germany

An engine control unit for motorsports applications needs to combine high performance, flexibility, and easy handling. In this talk we present a solution that integrates Model-Based Design, automatic code generation, FPGA technology and a powerful microprocessor. We will introduce the UNIT-Blockset, a custom embedded target for Real-Time Workshop. It supports different hardware architectures, while offering a common interface to the application. The functional model can be used on a range of ECU types without modifications.

We developed two additional tools to make handling easier. The DataManager handles the interface to calibration and measurement. It is a MATLAB GUI that configures the properties of Simulink objects exported to the ASAP2 file. DocGen produces the system documentation right out of the Simulink model during the build process.

Automatic Engine Control Code Generation with Integrated Automatic Static Code Verification
Robert Harmon, Chris Hote, Cummins - USA

The software design and development of engine controllers evolved into a product line architecture in order to keep up with increasing customer need for additional and reliable functionality delivered at lower cost and in shorter timeframes. The amount of embedded code for Cummins engines has increased ten-fold over the last decade.

Traditional engine controller module development methods cannot accommodate these increasing market demands. This is because traditional methods, which involve multiple manual interactions including hand coding, code review, and unit/integration tests, tend to be error-prone and time-consuming.

Given product line architecture, the viable approach is to use Model-Based Design for creating executable specifications, automatically generating controller code, and performing verification and validation (V&V) activities on the model. Furthermore, recent static analysis techniques are applied to generated code for static verification (such as runtime error detection and MISRA C: 2004 compliance) either very early in the software design process to find software reliability breaches before functional tests are performed, or later on for final sanity check.

The technology supports links back to the model and applies to different levels of the model, including subsystems and blocks, making it easy to resolve issues and perform traceability analysis. This paper describes a working example of an integrated environment using Model-Based Design with automatic code generation and advanced code analysis tools.

Modeling and Simulation of Distributed Automotive Systems with Simulink®
Dr. Thomas Ringler, Daimler Chrysler – Germany

Model-Based Design, simulation and automatic code generation have reached automotive series development. Within DaimlerChrysler Model-Based Design is used for central ECUs in the body domain. Currently each function (like seat heating, interior light control etc.) is modelled individually by means of a Simulink model.
Conclusion:

SimEvents provide a very general approach for modeling communication systems. Nevertheless the toolbox is quite powerful. The few blocks allow to model quite a huge variety of problems.
- Modeling real systems, explicitly modeling CAN-ID by servers is not really applicable and maintainable, as used in this case-study. Some other modeling constructs have to be found.
The intended uses-cases of CAN communication could be realized by using the toolbox. However applying the toolbox to automotive CAN applications additional features could help to make modeling more convenient.
- Currently for attributes only scalars can be applied. For detailed modeling application signals to be send via the CAN-Bus, Simulink busses should be allowed as attributes.
- Enhance Import of Communication description matrices
- Enhanced Datablock using SL-Buss Objects
SimEvents allows to model the communication aspects of distributed functions. However additional means are still needed to model the other central mechanisms of communication like network management, which allows to model the startup and shutdown of nodes within a communication network.
The ability of having in and out ports on the same side of a subsystem, helps cleaning up models. This concept is very helpful and should also be applied to native Simulink.
Enhancement to simulate functional behavior using the AUTOSAR VFB / RTE

DENSO's Model-Based Design Capability to Contribute OEM's Success
Manji Suzuki, DENSO – Japan

The ECU software size got huge as vehicles had more and more functionality, then many OEMs tried to introduce MBD process to reduce development time and improve software quality as well. However, the problem is that ECU development process is across OEM and suppliers, then supplier's MBD capability is one of the key issue to win a success in introducing MBD process. DENSO has started preparing for MBD process since 1996 to meet future expectation from OEMs, and have been accumulated knowledge and technology for nearly 10 years.

Model-Based Digital Filter Design for Improvement of LIDAR Sensors
Dr. Joachim Waßmuth, Dr. Bernd Lichte, Hella - Germany

At present advanced driving assistance systems (ADAS) represent one of the most innovative branches of modern vehicle development. Safety as well as comfort of the driver are the common aim of both OEMs and suppliers. Multi-beam infrared sensors like the Hella IDIS are well suited to realize functions like adaptive cruise control, precrash or automatic emergency brake, being less expensive in comparison to long-range radar sensors. While multi-beam infrared sensors deliver significantly better results for the lateral position and the width of objects, long-range radar sensors measure the relative velocity of objects directly using the doppler effect.

It is described how this disadvantage could be eliminated by methods of applied digital signal processing and model based design. Different filter algorithms were developed by using MATLAB and Simulink. As the algorithms have been implemented on an ECU using a 16bit microcontroller target, several aspects of fixed point arithmetic had to be considered. The model based design of the algorithms i.e. modelling signalflow diagrams of the filter algorithms was done in Simulink by using so called wave digital filter structures instead of classical direct form structures. These filters have several advantages like low coefficient sensitivity and garantueed stability under nonlinear operation conditions like rounding, thus ensuring system safety. The fixed point capabilites of both MATLAB and Simulink are ideal for simulation and analysis of this behaviour. It is important for the functionality, that the transient behaviour of the filters can be influenced, when the filter is called the first time. This transient behaviour is not part of the classical filter design process, where only the stationary behaviour is taken into account. For this reason initialization procedures for the filters have been developed.

The complete algorithms can be stimulated and tested with measured data, received via CAN, or with stimuli generated by a signal model, being implemented in Simulink as well. Last step of the design was the implementation of the algorithms on 16-bit microcontroller (!) by using the code generation capabilities of the Real-Time Workshop Embedded Coder.

Design and Calibration of the New Jaguar XK Adaptive Cruise Control System
Tim Jagger, Jaguar Cars - Great Britain

Jaguar first introduced Adaptive Cruise Control (ACC) on the XK8, in 1999. The new 2006 Jaguar XK features the next generation of Jaguar Adaptive Cruise Control. The new XK has been developed to tight deadlines and with significantly fewer prototype vehicles, than previous ACC projects. These constraints have driven Jaguar engineers to embrace new techniques for designing and calibrating the ACC system. Computer simulation, of the ACC system, has reduced the design and calibration time required to meet target performance requirements.

Development of the original ACC system followed a traditional iterative vehicle based route. There was neither the luxury of time nor prototype vehicles to follow a similar route on the new XK. Instead a new development method was employed – model based calibration. This paper will look at the use of Simulink models to calibrate the ACC system on the new Jaguar XK. Specifically it will look at creation of an ACC whole vehicle model, correlation of this model and calibration in the Simulink environment. A mathematical model was created in Simulink which captured the dynamics of the ACC radar, ACC control algorithm, engine management system, brake system, driveline and vehicle dynamics; essentially this was a Simulink model of the X150 whole vehicle ACC system. MATLAB was used to process test data from prototype vehicles and format it for use in the ACC model. The Simulink model was correlated against the test data. The correlated model was used in design of experiment exercises to evaluate the effect of different calibration settings. Changes in the EMS and brake system calibrations can also effect ACC performance. These changes were quickly evaluated in the Simulink ACC system model. Tests sessions on vehicles then became an exercise in validating calibrations derived in Simulink. Use of model based calibration techniques allowed Jaguar engineers to significantly reduce the number of design iterations and time required to develop ACC. Use of MATLAB and Simulink has typically allowed Jaguar to reduce a month of experimental test work into five days of simulation.

In summary this paper will focus on:

Development of an Adaptive Cruise Control vehicle model in Simulink
Correlation of the model against test data processed in MATLAB
Use of the model in to derive calibrations and assess against defined targets
Timescale reduction achieved through use of Simulink/MATLAB models

Development of a Diagnostics Tool for Fuel Cell Vehicles
Christof Nitsche, Mercedes Benz Technology - Germany

Fuel cell vehicles in general require different diagnostic approaches than conventional vehicles powered by internal combustion engines. Similar tough is the use of a CAN-Bus as a communication media. The combination of Vector-Informatiks Canape Software and MATLAB, Simulink, and Real-Time Workshop from The Mathworks provide the necessary freedom for the implementation of various diagnostic approaches for a fuel cell propulsion system. This paper will talk about how a set of fuel cell models (Fig. 1 middle), created by MATLAB m-scripts, can be implemented into Simulink (Fig. 1 lower left corner). This Simulink implementation can, by utilizing Real-Time Workshop, further be implemented into Canape for either online real-time modelbased comparison (Fig. 1 lower right corner) of simulated to real fuel cell system operating parameters (such as electrical current and voltage) or for an offline comparison for the detection of fuel cell system faults. In addition Canape allows the implementation of displaying and processing the individual cell voltages of the fuel cell stack, which is composed of several hundred cells (Fig. 1 upper left corner). The combination of these mentioned functionalities provides a useful tool for research and development.

Model-Based Development Applied to Electronic Architecture System Design
Eric Zink, PSA Peugeot Citroën – France

The number of distributed automotive systems is growing at an accelerating rate. To cope with the complexity of such systems in chassis control and power train domains, PSA Peugeot Citroën has adapted its methodologies and tools.

We will present our new development platform named "OSCAR" which integrates model-based technologies for electronic architecture design.

How Renault Is Using MATLAB®/Simulink® and Real-Time Workshop® for Vehicle Dynamics HIL Validation
Philippe Tchilinguirian and Alex Yvart, Renault – France

Many chassis control systems have appeared these last years and forced the automotive engineers to adapt their validation methodology.

It's not enough anymore to master the different conception tools, some of them for chassis tune-up, the orthers for ECU's validation.

From now on, modern cars tuning requires the simultaneous use of both tools along the whole life cycle. These new challenges have made Renault reconsider the engineering of MADA (which is the company's vehicle model) to perform full compatibility with dSPACE environment (used for ECU's validation). Using MATLAB/Simulink and Real-Time Workshop, the MADA's architecture was re-built, and allows the same vehicle model to be used as well for chassis and ECU's validation.

We aim to present the technical issues of this work.

TNO Delft-Tyre and SimMechanics: Benefits and Examples of Vehicle-Tyre-Road Physical Modelling
Lennard Verhoeff, TNO – The Netherlands

Each design stage and application imposes different requirements on vehicle-tyre-road interaction simulation. SimMechanics and the TNO tyre model MF-Swift enable high-fidelity physical modeling of vehicles and vehicle systems from concept to validation. MF-Swift provides user-selectable detail in the tyre-road interaction to support applications such as global chassis control development, vehicle handling and ride comfort studies or hardware-in-the-loop simulations. The benefits of SimMechanics in combination with MF-Swift are explained using examples of motorcycle, race car and truck simulation studies.

Converting Legacy Embedded Control Software to Executable Specifications
Koichi Ueda, Toyota - Japan

This presentation describes some key aspects of the conversion process and discusses some important challenges in having an efficient process which yields the correct model. This conversion process is adopted for the large scale conversion task, and the result of this work improves the quality of the software and the efficiency of the development, in other words, leads to Model Based Development.

Development of an Automotive Real-Time Simulator for Preventive Safety Studies
Alessandro Malvasi, University of Pisa - Italy

Concerning the field of preventive safety, recent studies showed an increasing attention towards the monitoring of the psycho-physical state of a vehicle driver during time, in order to be able to alert a sleepy driver in advance.

In some interesting papers it is discussed how a sleepy driver can be recognised on the basis of characteristic signals identifying the "vehicle system" and in particular the steering system [1, 2, 3, 4]; indeed, in this way it seems possible to develop an economic and non-intrusive system, although intrinsically less precise, with respect to a system based on drivers physiological parameters.

For research purposes in this field, at the Department of Mechanical, Nuclear, and Production Engineering of the University of Pisa, an interactive driving simulator is at an advanced state of development; driving simulators appear the best solution for preliminary and safe studies [5, 6, 7, 8, 9, 10]. The simulator will allow to carry out tests during which physiological parameters together with signals related to vehicle motion (displacement, velocity and accelerations components) and drivers actions (steer angle and torque) will be recorded simultaneously. In this way, it will be possible to analyse any relationship between the different quantities and to realize if the signals describing the vehicle motion conditions themselves can be used to recognise potentially dangerous situations.

In this paper the dynamic model of the vehicle, developed in the MATLAB/Simulink environment, and the way this model is managed in real-time using xPC Target and Real-Time Workshop are discussed; the model is interfaced with a real steering system and instrumented pedals; vehicle spatial position and orientation, obtained by the vehicle model, are used by a graphical system aimed at updating the simulation scenario.

The non-linear vehicle dynamic model has 14 degrees of freedom and is based on [11]; the main vehicle’s subsystems, such as engine, brakes and tyres were modelled; about tyres, the non standard modelling for low speed calculations described in [12] has been implemented.

For data acquisition from the steering system a CAN network with three nodes was set up; on the other hand signals from pedals are acquired using a National Instrument data acquisition board.

The UDP communication was chosen to manage output signals, in this way audio and video software get signals directly from the network, without using MATLAB.

Results of some preliminary tests, carried out for both system validation and in order to evaluate system performances are also presented and discussed in the paper.

Standardized Software Modules for Model-Based Design Workflow
Dr. Oliver Schütte, Eva Kalix, WABCO - Germany

In this paper a layout structure for Simulink models is presented that standardizes and eases the handling of software modules for multi-user- and multi-variant projects.

Regarding the overall model design the data exchange between modules is based on a standardized bus infrastructure. Each module is embedded in an interface subsystem that connects to a model data bus, to a parameter bus, and to a monitoring bus, which may be used for test reporting or visualization.

The data bus is collecting and dispensing the runtime data of the module interfaces. The parameter bus carries the model initialization constants and is created automatically from object oriented m-structures that were defined for each module.

Looking at software quality modules or software components using this bus-interface structure can be easily (and automatically) assembled in case a different variant or revision is needed. Integration, e.g. into different test harnesses providing stimuli for module or component specific data bus signals, is also possible. If a new variant configuration has been constructed badly, bus selectors will send a distinct error message in case a module misses a signal.

Concerning the overall workflow a set of m-scripts was created and is outlined to initialize MATLAB/Simulink, to construct the desired variant and revision of software modules and to manage configuration information, to register the present modules and its corresponding parameter files, to check the modules for extra company design rules and subsystem interfaces, to invoke Real-Time Workshop or autocode generation procedures, in an accurately defined way.

The overall design of a software project usually is available at project start including modules and interfaces. Therefore, a basic but executable model structure can be build up in the very beginning of a project and will be filled by and by as recommended in SPICE driven development processes.

The separation of the functional model from its parameter data as well as the clearness of the model structure outweighs the higher wiring effort of a bus layout. Standardized user m-scripts for workflow management, variants handling and build processes guarantee a defined output of the tool chain.

Mathworks Presentations

Talks from The Mathworks will include:

Release 2006a enhancements to MATLAB^® and Simulink^®
Advanced ECU code generation technologies and popular production deployment workflows
The latest addition to the physical modeling product set
Verification, validation, and test solutions

Master Classes

Running concurrently with the general session track each day are three 90-minute master classes. These sessions are for advanced users who want to gain deeper knowledge of MathWorks tool use and capabilities. Attendees are eligible to register for one master class session each day.

Pre-registration is required for these sessions because seating is limited. Onsite registration for these sessions will be based on space availability.

MAAB Style Guide for 2006
Production Code Generation
Advanced Programming Techniques in MATLAB^®
Modeling of Mechanical, Electrical, and Hydraulic Systems in Simulink^®

Products & Services	Industries	Academia	Support	User Community	Company