Current mode detection in hard real-time automotive applications dedicated to many-core platforms

This paper proposes a technique for determining the current mode in an electronic control unit (ECU) during run-time. We use a decision tree classifier which observes the latest execution times of processes (runnables). When a mode change is detected, the migration of runnables is performed to decrease the number of active cores leading to considerable energy savings while still not violating any of timing constraints. The proposed approach consists of both off-line and on-line steps, whereas more computational intensive steps are performed statically. In the presented automotive use case, the current mode is detected with 100% accuracy while observing execution time of a particular single runnable. The migration time of systems with dynamic mode detection based on the runnable execution time with various periods is also provided.