KI4BoardNet

KI4BoardNet explores new concepts for resource-optimized, continuously learning, distributed power management systems. This includes methods for developing application-specific system-on-a-chip (SoC) architectures for hardware-assisted, accelerated execution of machine learning algorithms. For this purpose, a generator-based design process for RISC-V systems will be adapted to allow for flexible SoC architecture configuration and optimization.

Regarding the generated SoC platform, different adopted model architectures for AI-supported energy consumption prediction and their integration into a distributed onboard network load management will be researched. The project will also examine novel algorithms based on auction theory for resource efficiency and economical use of available energy within the board network. An important aspect of the planned system is its ability to learn continuously and thus continuously adapt to various usage scenarios.

In addition to voltage stabilization in future board networks; the research results will enable the implementation of new AI applications.