Edge Computing for Local LLM-based HMI
Large Language Models (LLMs) are currently a very popular research topic. Our goal is to transfer current developments to real robotic systems without internet connection. This will improve the possibilities of human-machine interaction.
As part of the bachelor thesis, a sub-project will be carried out. The first step will be a literature review on current methods for speech-interactive local integration of an LLM. In other words, which methods, tools and models are suitable and what computing power is required for an acceptable runtime for user-friendly interaction. High performance computers and also embedded/edge computing devices are available for the evaluation. Subcomponents of the framework under consideration include speech-to-text, LLM and text-to-speech.
Advisor:
- Sven Lange, sven.lange@…
Requirements:
- Basic knowledge in Linux/Ubuntu would be good.