Use Case 1 · Performance
Identifying the Most Promising Planner
Challenge
Picking a planner for a new domain is largely trial-and-error across many planner / heuristic combinations.
Ontology-driven approach
1. Encode IPC-2011 performance as a knowledge graph.
2. Query for the best-performing planner per domain.
3. Apply that ontology policy instead of a random choice.
| Policy | Nodes expanded | Success rate |
|---|---|---|
| Ontology policy | Fewer | Higher |
| Random policy | More | Lower |
Ontology-guided selection expands fewer nodes and solves more instances than a random policy (IPC-2011).
Bonus: the ontology also yields macros — action-ordering constraints that further speed up planning.