Cognition · Software component

Learned Heuristic Estimator

Software componentCognitionCognition & Memoryarc:LearnedHeuristicEstimator

A heuristic estimator that runs a trained neural network over state features (e.g., obstacle configuration) to predict true remaining path cost.

Responsibility. Predicts near-exact remaining cost per node via learned-model inference.

Also known as: Neural network heuristic

Variant of Heuristic Estimator abstract

When to choose. Choose only when inference is cheap relative to node expansion cost, so higher accuracy reduces total search time.

Direct neighbourhood (hover for relationship types)

Relationships

hosts structural

Learned Heuristic Model Ch5.6

alternative to variability

Geometric Distance Heuristic Ch5.6

Design guidance

MUST justify its per-evaluation inference cost against the node expansions it saves before replacing a cheap geometric heuristic.

Quantitative guidance

As stated by the sources; verify before use.

Learned heuristic within 2% of h*(n) cut expansions 380 -> 120 (68%) but 15 ms per inference made heuristic time 1,800 ms, far exceeding the 45 ms saved (Ch5.6).

Classification

Patterns: Learned heuristic
Quality attributes: Functional suitability: correctness and validity (ISO/IEC 25010 | NIST AI RMF: valid)

Sources

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.