INPROVF: Leveraging Large Language Models to Repair High-level Robot Controllers from Assumption Violations

2025-03-17

AI Summary

High-order robot controllers often fail when environmental assumptions are violated, and existing formal repair methods suffer from prohibitive computational cost and poor scalability to large state spaces. Method: We propose an LLM-guided iterative automated repair framework that translates symbolic models into natural language prompts and leverages formal verification feedback to steer large language models in generating and refining repair strategies. Contribution/Results: This work pioneers the integration of LLMs with formal verification for controller repair. Evaluated across 12 multi-scale scenarios—spanning diverse state space sizes and task complexities—the framework achieves efficient, scalable assumption-violation repair. Experiments demonstrate substantial improvements in repair efficiency over purely formal approaches, while supporting more complex controller architectures and dynamic environment modeling.

Technology Category

• Planning, Routing, and Scheduling: Replanning and Plan Repair
• Intelligent Robots: State Estimation
• Natural Language Processing: Safety and Robustness

Application Category

• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Search Tool Learning with LLM: Teaching LLMs to invoke search and make use of retrieved information

Abstract

This paper presents INPROVF, an automatic framework that combines large language models (LLMs) and formal methods to speed up the repair process of high-level robot controllers. Previous approaches based solely on formal methods are computationally expensive and cannot scale to large state spaces. In contrast, INPROVF uses LLMs to generate repair candidates, and formal methods to verify their correctness. To improve the quality of these candidates, our framework first translates the symbolic representations of the environment and controllers into natural language descriptions. If a candidate fails the verification, INPROVF provides feedback on potential unsafe behaviors or unsatisfied tasks, and iteratively prompts LLMs to generate improved solutions. We demonstrate the effectiveness of INPROVF through 12 violations with various workspaces, tasks, and state space sizes.

Problem

• Repair high-level robot controllers using LLMs and formal methods.
• Overcome computational expense of formal methods in large state spaces.
• Iteratively improve repair candidates with feedback and verification.

Innovation

• Combines LLMs and formal methods for repairs
• Translates symbolic data into natural language
• Iteratively improves repair candidates via feedback