An Efficient Hierarchical Scheduling Framework for the Automotive Domain

Modern real-time systems have become exceedingly complex. A typical car is controlled by over 100 million lines of code executing on close to 100 Electronic Control Units (ECU). With more and more functions being implemented in software, the traditional approach of implementing each function (such as engine control, ABS, windows control) on a dedicated ECU is no longer viable, due to increased manufacturing costs, weight, power consumption, and decreased reliability and serviceability (Nolte et al., 2009). With the ECUs having increasingly more processing power, it has become feasible to integrate several functions on a single ECU. However, this introduces the challenge of supporting independent and concurrent development and analysis of individual functions which are later to be integrated on a shared platform. A popular approach in the industry and literature is component based engineering, where the complete system is divided into smaller software components which can be developed independently. The Automotive Open System Architecture (AUTOSAR) (AUTOSAR, 2011) standard is an example of such an approach in the automotive domain. It relies on a formal specification of component interfaces to verify the functional properties of their composition. Many functions in automotive systems, however, also have real-time constraints, meaning that their correct behavior is not only dependent on their functional correctness but also their temporal correctness. AUTOSAR does not provide temporal isolation between components. Verifying the temporal properties of an integrated system requires complete knowledge of all functions comprising the components mapped to the same ECU, and therefore violates the requirement for independent development and analysis.