LLM-Assisted Architecture Recovery for ROS 2 Systems

A new approach uses large language models (LLMs) to recover hierarchical structural architecture from ROS 2-based robotic systems. The method extends a previous blueprint-guided pipeline with refined prompting and a staged recovery strategy based on multi-level intermediate architectural representations. This addresses the challenge that structural decomposition and integration semantics are often implicitly encoded across distributed artifacts like source code and launch files. Existing techniques focus on node-level entities and communication wiring, but fail to recover hierarchical structure across multiple abstraction levels. The work is described in arXiv paper 2605.20055.