Future Work
Where This Goes Next
🧠
Language Models & Ontologies
- Ground LLM plan explanations in the ontology for context-aware, faithful narratives.
- Natural-language SPARQL querying over planning knowledge graphs.
- Template learning to adapt explanations to the user.
🔗
Neuro-Symbolic Representations
- Integrate symbolic ontology knowledge with neural planners & learners.
- Use structured constraints (e.g., macros) to guide and explain learned policies.
- Safety-rule checking & conflict-driven diagnostics.
Also: extend maPO to dynamic, real-world multi-agent & robotics settings (via SOSA/SSN), richer conflict taxonomies, and temporal / numeric planning.