LLM-Assisted Analysis of Architectural Evolution in Distributed Robotic Systems

Robotic systems are increasingly built as distributed software architectures, where perception, planning, and control components are deployed as loosely coupled modules (e.g., ROS nodes). As robots become more autonomous, these architectures evolve rapidly, often without explicit architectural documentation or systematic design governance. This leads to architectural drift, hidden dependencies, and maintainability challenges that are difficult to detect using traditional static analysis alone.

This project explores how Artificial Intelligence for Software Engineering (AI4SE)—in particular Large Language Models (LLMs)—can assist developers and researchers in understanding, analyzing, and tracking the evolution of software architectures in distributed robotic systems. Robotics software serves as a realistic and challenging domain for AI-assisted architectural analysis due to its distributed nature, configuration complexity, and tight coupling with physical components.

The student will combine classical software architecture analysis with LLM-based techniques to extract architectural information, analyze software evolution, and explain architectural changes over time. The focus is on software engineering tasks, not robot control or hardware design.

Project contributions

This project makes the following contributions:

• An empirical study of architectural evolution in distributed robotics software systems.
• An LLM-assisted approach for architectural recovery and dependency analysis in ROS-based systems.
• Techniques for detecting and explaining architectural drift, erosion, and modularity degradation.
• A prototype tool that integrates architectural extraction with LLM-based summarization and explanation.
• Empirical insights into the strengths and limitations of LLMs for software architecture analysis.

Available spots

Available spots: 2

Pointers to literature

• Hou, X. et al. Large Language Models for Software Engineering: A Systematic Literature Review. ACM Transactions on Software Engineering and Methodology (TOSEM), 2024.
• Chen, Z. et al. How Well Do Large Language Models Understand Software Architecture? Proceedings of the International Conference on Software Engineering (ICSE), 2024.
• Timperley, C. et al. ROBUST: A Dataset of Real Bugs in the Robot Operating System. Empirical Software Engineering, 2024.
• Le Goues, C. et al. Software Engineering for Autonomous Systems. IEEE Software, 2023.
• Brugali, D. et al. Software Engineering for Robotics. IEEE Robotics & Automation Magazine, 2022.