Cognition · Software component

Graph Search Planner

Software componentCognitionCognition & MemoryVariation point (abstract)arc:GraphSearchPlanner

A task planner that finds a lowest-cost action or path sequence by systematically searching a weighted state-space graph from the current state to a goal state.

Responsibility. Systematically searches a state-space graph for a lowest-cost path to the goal.

Also known as: Pathfinding planner, Search-based planner, Informed search planner

Variant of Task Planner abstract

Direct neighbourhood (hover for relationship types)

Variants

Variant	When to choose
Bounded-Suboptimal Search Planner	Choose for real-time games, robots under time pressure and interactive systems where responsiveness outweighs optimality and cost may exceed optimal by up to factor w.
Memory-Bounded Search Planner	Choose when memory is scarcer than time: embedded systems, mobile devices with strict memory budgets, or search spaces far exceeding available RAM.
Optimal Heuristic Search Planner	Choose when the state space admits informative admissible heuristics, optimal solutions matter more than computational efficiency, the branching factor is manageable and transitions are deterministic (e.g., route planning on a known warehouse floor plan).
Uniform-Cost Search Planner	Choose when many destinations share one origin (20+), when the graph is tiny (~50 nodes), when repeated queries on a static graph amortise preprocessing, or when no meaningful goal-distance heuristic exists.

Relationships

invokes dependency

Heuristic Estimator abstract Ch5.6

is invoked by dependency

reads dependency

State-Space Graph Ch5.6

produces lifecycle

Execution Plan abstract Ch5.6

Design guidance

SHOULD be chosen for deterministic transitions with well-defined costs and a clear goal direction; stochastic or adversarial domains favour Monte Carlo Tree Search.
MUST treat its search results as stale when edge costs or obstacles change during execution and hand over to a replanner.
SHOULD select the search variant from time constraints, solution-quality tolerance, memory budget and number of destinations.

Quantitative guidance

As stated by the sources; verify before use.

Warehouse robot: optimal 47 m path required expanding 380 of 50,000 nodes in 185 ms, exceeding a 100 ms path-update deadline (Ch5.6).

Classification

Patterns: A* searchBest-first searchOpen/closed list searchf(n) = g(n) + h(n) evaluation
Quality attributes: Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)Performance efficiency (ISO/IEC 25010)
Risks mitigated: Exhaustive exploration of irrelevant paths

Sources

Ch5.5: T. Nguyen, "Monte Carlo Tree Search Fundamentals," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.5. ISBN: 9798244538229.
Ch5.6: T. Nguyen, "A* Search and Replaning," in Mastering Agentic AI Systems: Guide for the NVIDIA NCP-AAI Exam, 1st ed. 2026, ch. 5.6. ISBN: 9798244538229.