Source record

Published by arXiv cs.RO (T1 source). The original is at https://arxiv.org/abs/2607.15674.

Pipeline notes

The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.

The paper introduces GPSFSM, a neurosymbolic framework that combines Large Language Models with Finite State Machines for robot behavior planning. It includes Fabric (an FSM engine) and PromptTools (a ROS 2 LLM interface). Experiments show higher plan-generation success than BTGenBot, especially in zero-shot scenarios. Open-source ROS 2 stack released.

First generative FSM framework for robotics with released open-source stack; relevant to researchers building LLM-driven robot behavior planners, though narrow outside that domain.

Cited by

No citations on record.