Towards Model-Based Generation and Optimization of AUTOSAR Runnable-to-Task Mapping
Thomas E. Wilhelm, Raphael Weber · 2019
In this paper we present work towards automating two process steps supporting the optimization of the runnable-to-task mapping in automotive multi-core control units. We describe these steps in close relation to the AUTOSAR methodology to facilitate the integration with existing design processes. The first step is the automated generation of an initial configuration that balances the core utilization using constraint programming. The second step is the optimization of an existing configuration based on dynamic system behavior using an evolutionary algorithm. An abstract intermediate representation provides interoperability with existing AUTOSAR tools. We use a small case study to evaluate the feasibility of our approach.