There are many economic and technical arguments for the reduction of the number of Electronic Control Units (ECUs) aboard a car. One of the key obstacles to achieve this goal is the limited composability, fault isolation and error containment of today’s singleprocessor architectures. However, significant changes in the chip architecture are taking place in order to manage the synchronization, energy dissipation, and fault-handling requirements of emerging billion transistor SoCs (systems-on-a-chip). The single processor architecture is replaced by multi-core SoCs that communicate via networks-on-chip (NoC). These emerging multi-core SoCs provide an ideal execution environment for the integration of multiple automotive ECUs into a single SoC. This paper presents a model-based software development method for designing applications using these multi-core SoCs. 1.

Modern automotive electronic control systems are distributed, networked embedded systems. Diagnostic routines implemented on individual components cannot adequately identify the true cause of anomalous behavior because their view is restricted to component-local information. A growing trend in diagnostics research for these systems is to use system-level approaches to diagnose anomalous behavior and provide a consistent, global view of the system’s health. Current approaches are typically motivated by a desire to improve either off-line maintenance or run-time safety. Automotive systems are safety-critical and are required to be highly fault-tolerant in operation. Automotive systems consist of mechanical, hydraulic, software and hardware components. There is a staggering amount of embedded computing within automotive systems. For instance, current GM vehicles contain dozens of microprocessors and dozens of megabytes of software.