specification of the subsystems, can specify the requirements in a semiformal language such as the Unified Modeling Language (UML). The overall behavior (functional net- work) and architecture netlist of the distrib- uted system constitutes the output of this specification phase.
This methodology supports the entire auto- motive electronic design chain. Carmakers specify the overall system to the first-tier sub- system suppliers, who develop algorithms and implementations in software and hardware platforms. The second-tier suppliers, semi- conductor and software companies, develop components of the hardware platform and software. Carmakers use the virtual platform of the overall electronic system to validate their partitioning and specification, and subsystem suppliers use it to ensure that their compo- nents work as expected when integrated.
In this methodolog , the model-based soft- ware design approach is important. Through synthesis export, a model described in func- tional terms migrates to an efficient software implementation. The subsystem supplier develops most of the software, but carmakers are developing an increasing amount of criti- cal software and can benefit substantially from this approach.
The new methodology’s key distinctions are its use of the following:
Virtual platform. A virtual platform sup- ports system testing and prototyping (hardware-software architecture) via simulation.
Virtual application models. Virtual mod- els of the application software and the target hardware-software architecture (bus controllers, CPUs, RTOS sched- ulers, and communication protocols) let designers create a virtual prototype of the entire distributed application. Designers import the application software models from previous designs or write new soft- ware for the system under development. Designers develop the architectural mod- els within the virtual models (for exam- ple, a communication protocol model), using a standard C++ application pro- gramming interface.
Other virtual models. Besides virtual application models, virtual models of the
environment, of complex human- machine interactions, and of test bench- es that provide stimuli to the system under test are also necessary. Designers either import these models from tools such as MathWorks Simulink or author them within the system.
The new methodology makes a major shift to an integrated design style in which design- ers model the entire ECU network along with the application and base software used to cus- tomize the platform for a particular car series.
Integration takes place at the virtual level. Automatic configuration of tools for protocol analysis and implementation is based on the results of simulations of the virtual model. For example, once a designer has decided how to distribute the pool of functions among the ECUs, a downstream code generation tool can use this information (number of tasks needed, scheduling policies, and so on) to generate the RTOS scheduler. At the same time, designers configure the downstream tools for commu- nication protocol analysis, using the configu- ration data determined at the virtual level (type of protocol, frame packaging, commu- nication cycle, and redundancy management policies). Thus, a step that is currently man- ual or that requires intensive user intervention (for example, the designer must explicitly specify messages sent over the network bus) is automatic in the new flow.
In the new methodolog , timing estimation takes place in the earliest design phases, before implementation. Typical estimates are for soft- ware task execution times and network com- munication latencies. These estimates might shorten algorithm and platform (single ECU or network) exploration considerably. Estima- tion models can be provided by the subsystem supplier or developed by the system designer for use as a specification to the supplier.
Platform-based design paradigm
Figure 3 depicts the core of the new methodology as an instance of platform-based design, a paradigm shift in design, verifica- tion, and testing. The platform-based design paradigm is a meet-in-the-middle approach. It leverages the power of top-down methods and the efficiency of bottom-up methods. The