Automated Context Adaption of Iterative Software in Constraint Execution Environments

Context

In the realm of autonomous driving, embedded software components and algorithms, such as route planning, undergo iterative execution. The resulting trajectory is updated at each time step, and the continuous execution at shorter intervals contributes to enhanced result quality. This project aims to construct a scheduling model that dynamically adjusts the expected accuracy based on a specified utilization and context. A context under which the utilization is constraint should lead to an execution with less accuracy while simultaneously freeing the processor to do other work. In every case a given minimum accuracy shall remain.

Targets

The aim of this thesis is to implement a scheduler model to dynamically adapt iteratively executed tasks to reduce the quality of the result in some iterations to free up processor time for other tasks.

Requirements

• Solid mathematical foundation for modeling real-world scenarios and algorithms within a research context.
• In-depth understanding of the architecture of Embedded Multi-Processor System-on-Chips (MPSoCs).
• Proficient knowledge in Operating Systems and in C/C++.