Code generation from Amalthea to Linux and FreeRTOS on Heterogeneous Mixed-Criticality System-on-Chips (HeMCSoCs)

Sammanfattning

Modern automotive and cyber-physical systems are increasingly required to execute a wide range of tasks with varying levels of criticality, including advanced driver assistance, real-time control, and onboard artificial intelligence. These demands require heterogeneous platforms to provide strong isolation, predictable timing behavior, and fault containment. The Carfield platform, based on the open RISC-V instruction set architecture, addresses this by partitioning execution across specialized domains for safety, security, general-purpose processing, and hardware acceleration. Despite these capabilities, developing and deploying mixed-criticality applications on such systems remains complex, due to challenges in coordinating software across different operating systems and managing access to shared hardware resources. This thesis explores the application of Model-Based Design (MBD) using the Amalthea modeling framework to simplify software development for the Carfield platform. The work includes replicating the Carfield setup and extending its toolchain with an automated code generation framework. A model-based design flow is proposed, allowing Amalthea models to specify functional and non-functional system properties, which are then translated into template code targeting Linux and FreeRTOS operating systems. The generated code supports real-time task offloading and configuration of hardware mechanisms such as reservation-based AXI interconnects. Experimental validation confirms that functional and non-functional properties configured in the Amalthea model are correctly translated into the template code, and the system behaves as expected.