Research repository based on Microsoft's recent study; comprises findings on large language model-enhanced visual simultaneous localization and mapping for robotics tasks and motion planning.
The basic structure of the project covers several LLM integrations on Microsoft's robotics simulation AirSim. Logical reasoning, common-sense capacity, and aerial navigation capabilities are considered evaluation metrics of LLM performances.
The project is under development at this time; the source code may be incomplete and could be updated occasionally.