PDF-Flyer

Concepts for Self-Adaptive Automotive Network Architectures

Prof. Dr.-Ing. Jürgen Teich, Universität Erlangen

Today’s embedded systems, especially in the automotive area, are typically distributed. Here, problems of integration, reliability, and time-varying applications and user demands require to rethink the design and integration of hardware and software. A major requirement to enhance integration and extendability is adaptivity. Existing methodologies that optimize the partitioning
of computational tasks to hardware and software at compile time become obsolete or inefficient in this context as the optimal use of existing resources cannot be foreseen. Here, we investigate novel strategies that solve the problem of hardware/software partitioning online: Once there are changing computational demands, the system should be able to dynamically adapt the assignment of functionality implemented in hardware or software to network nodes. Moreover, online hardware/software partitioning can be applied to reestablish functionality in case of node or link defects at no or very low additional cost, and thus provide fault-tolerance.

As an example, we present a demonstrator network called ReCoNets which is based on reconfigurable devices such as FPGAs together with internal CPU cores. With this network and the corresponding operating infrastructure, it is possible to assign tasks implemented in either hardware or software, dynamically to the resources in the network without global control or interaction.

The application of a fault-tolerant driver-assistance system is used as a case study of the benefits of self-adaptive networking in the car.

[zurück]