New: PPU - a scalable SIMD vector DSP for cross-application use cases
Performance requirements vary across different future automotive applications. This is challenging existing embedded microcontroller products and raises a need for new SoC architectures.
ARC EV based Parallel Processing Unit (PPU) supports meeting new performance requirements arising from eMobility, automated driving / advanced driver assistance systems (ADAS), and new E/E architectures in the increasingly connected future vehicles. It has a vector DSP architecture to enable mathematical modeling of complex heterogeneous systems.
The main applications in eMobility will be in on-board charger and DC/DC converters, traction motor inverters, or battery management systems. In ADAS, PPU will support new use cases in radar, sensor fusion, or lidar. PPUs data processing capabilities will be an advantage for performing the tasks of an infrastructure controller for vehicle-to-x scenarios or enable AI-based implementations in future domain controllers.
PPU is a flexible architecture to address automotive applications with fast execution times and large data processing requirement
The implemented tasks and use cases will differ per application but two main clusters can be identified. On the one hand, PPU allows complex data processing and observer-based controlling of sensor actuator systems (e.g. for traction motor inverter or DC/DC converter control). On the other hand, it enables the implementation of artificial neuronal networks (MLP, RBF, RNN, CNN) based system modeling (e.g. for virtual sensors, state of health/state of charge optimization in battery management systems, and predictive vehicle motion control for in future domain or zone controllers) and object classification (e.g. Sensor Fusion) solutions.