Abstract
Keywords: design assistance, integration, semi-automation, timing, tools
Over the last years, significant effort has been spend to introduce model based software development methods and supporting tools in the development of electronic control units (ECU). One important future trend will be to complement model based development with smart assistance technologies. In the future, tools will support developers by offering assistance how to improve designs and calculate the impact of their design decisions.
For example, mapping functional building blocks on ECUs requires a sound understanding of multiple criteria, such as resource, performance, dependencies, cost, safety, and electrical power consumption. Providing assistance by presenting developers alternative choices and their respective impact on system performance, power consumption and cost will significantly support the decision making process. Mapping functionality to ECUs without having smart assistance in analyzing the impact of the mapping only addresses constructional aspects but lacks an analysis about the correctness of the execution behavior. Another example is the translation of requirements into design models. Assisting the developer in the creation of the design model and, at the same time, providing early feedback on the effect of design decisions on the implementation of initial requirements will significantly improve the quality of the design. In addition, through the early detection of design mistakes overall development time will be reduced.
How can such assistance be achieved? To illustrate the answer we sketch the assistance for optimizing the runtime performance of a system. Today, commercial tools exist to determine worst-case execution times through precise calculations for ECU software. Based on this data a system-level timing analysis, e.g., through holistic schedulability analysis, can be carried out for the SW on a single ECU or in a network of ECUs. The tools use analytical models to calculate the timing properties and the timing behavior. Tomorrow, this knowledge will be combined with suggestions on how to improve the timing behavior of ECU software. These will be design suggestions for architecture models as well as implementation suggestions for rather detailed models. Furthermore, the effects of alternative suggestions, such as using an event triggered approach instead of a polling approach, can be analyzed precisely and presented to the developer. For optimizing runtime performance, assistance will be achieved by providing an analytical model and effective patterns to improve timing behavior. Both parts constitute the timing assistance.
Future assistance technologies are "smart", because the assistance will be carried out semi-automated by offering developers choices within their concrete development task. There are many more fields of application for assistance, which include on/off-board partitioning, diagnostics, refactoring, variability, and safety. The major driver to complement tools with smart assistance is to erase design mistakes early, and to guide the developer to successfully master difficult tasks in system design and integration.