LLM-Based Assistance for Capability-Based Planning in Industry
A new hybrid assistance system combines an LLM layer with capability-based Satisfiability Modulo Theories (SMT) planning to improve interpretability and adaptability in industrial automation. The symbolic planner ensures formal correctness, while the LLM handles natural-language interaction, explanation, and adaptation. This addresses limitations of existing capability-based approaches, such as difficult-to-interpret solver feedback and the need for manual knowledge model updates. The system is detailed in arXiv:2605.28666.
Key facts
- Hybrid system augments capability-based SMT planning with an LLM layer.
- Symbolic planner maintains formal correctness.
- LLM layer enables natural-language interaction and explanation.
- Addresses issues of solver feedback interpretability and model adaptation.
- Designed for dynamic industrial environments with modular resources.
- Published on arXiv with ID 2605.28666.
- Focuses on automated planning of process sequences.
- Uses semantic knowledge models for resource functions.
Entities
Institutions
- arXiv