Conceptional design of an open framework for optimizing the distribution of hardware and software components in control networks for vehicles

Student:Christoph Neumann
Title:Conceptional design of an open framework for optimizing the distribution of hardware and software components in control networks for vehicles
Type:diploma thesis
Advisors:Kókai, G.; Hardung, B.
State:submitted on June 10, 2005

Existing middleware architectures for control networks in the automotive industry allow transparent relocation of software components, if boundary conditions, e.g. according to ROM/RAM size or CPU usage, are met. An exemplary size of such control networks is about 80 electronic control units, connected by 4 possible types of bus networks and hundreds of functions and software components running. Commonly the distribution of the components is generated by senior engineers and their expert knowledge, without any tool-support. Often it is not possible for them to optimize the distribution according to multiple objectives. Shortcomings are mostly recognized in a late construction phase, when the vehicles are already being built. In the future the relocation facilities shall be used for cost savings and the process shall be automated.
Task formulation:
The student has to develop a framework, that allows the application of optimization algorithms to the distribution of hardware and software components.The framework has to deliver the configuration of new location distributions of the components. Multiple objectives have to be considered as optimization criteria (like costs minimization or bus load balancing) as well as constraints (e.g. according to ROM and RAM size limitations or availability of hardware pins). Initial distribution configuration is either given, or has to be generated from scratch. Furthermore, partial configuration that need strict adherence have to be definable, that means that the location of some components is statically determinated. The most important tasks in this thesis are the conceptual design of the modular and configurable framework, and its prototypical implementation. The required input data has to be identified and for storing the information a database model has to be developed. The optimizing kernel is not primary goal. A basic random-based generation could be used. But the incorporation of common optimization algorithms, like evolutionary algorithms or swarm intelligence, has to be prepared. Therefore the student has to deal with the variety of these algorithms by creating a generic interface. The framework has to allow modular integration of different objectives, but the main one is costs minimization. Similarly, constraints have to be dealt with by a generic mechanism. The focused constraints are hardware availabilities, like ROM, RAM, pins and wire. To guarantee a broadness in consideration the student has to read up architecture description languages, like EAST EEA ADL.An XML import and export for the distribution configuration is desired, an Editor supporting the engineers in providing the hardware topologies et al. is not subject of this thesis.

watermark seal