Towards LLM-Assisted Architecture Recovery for Real-World ROS~2 Systems: An Agent-Based Multi-Level Approach to Hierarchical Structural Architecture Reconstruction
Researchers have developed a new method for recovering software architecture from complex ROS~2 systems using Large Language Models (LLMs). This approach refines LLM prompting for better consistency and introduces a staged recovery strategy with multi-level intermediate representations. The enhanced pipeline was evaluated on a challenging real-world robotic system, demonstrating improved structural consistency, scalability, and robustness in architecture reconstruction. AI
IMPACT Enhances software engineering practices for complex robotic systems by improving architecture recovery.