Expand description
Best-first backward-chaining planner over causal triples.
Starting from a goal state, finds triples where the object matches the goal and recursively plans to achieve the subject. Produces an ordered list of actions (triples to make true), expanding the lowest-cost subgoal first.
Cost estimation: each action’s cost is 1.0 / (1 + evidence_count) where
evidence_count is the number of triples supporting the action.
Structs§
- Plan
- A complete plan: an ordered sequence of actions to achieve a goal.
- Planned
Action - A single planned action: a triple that needs to be made true.
- Planner
- Backward-chaining planner over the triple store.