Automatic Finite-State Machine Generation for Robotic Tasks from Natural Language Specifications
Guided by Maya Cakmak
Finite State Machines (FSMs) are a common representation used in programming robots due to their expressivity and alignment with how robot tasks tend to be modularized. However creating complete, robust FSMs for different tasks is tedious and error-prone. Further, for any high level preference change, such as level of desired human assistance, many parts of an FSM needs to be re-written. In this project we will explore generation of robot FSMs from specifications, using large language models (LLMs).