Cognition · Software component
Contingency Planner
Software componentCognitionCognition & Memoryarc:ContingencyPlanner
A task planner that, at planning time, generates a primary plan plus pre-computed conditional branches for the most likely, high-impact failure modes.
Responsibility. Produces branching plans with prepared responses for anticipated failures.
Also known as: Deliberative replanner, Conditional planner
Variant of Task Planner abstract
Relationships
Design guidance
- SHOULD prepare a contingency only when failure probability x reactive delay exceeds the added planning overhead (P x D > C).
- SHOULD limit contingencies to the few failure modes covering most disruptions (typically 2-5) and to one or two contingency levels, leaving rare compound failures to reactive replanning.
- SHOULD derive failure frequencies from historical execution logs.
Quantitative guidance
As stated by the sources; verify before use.
- Branching plan took ~200 ms vs 50 ms for a linear plan but removed 2-3 s of replanning delay per empty-bin event (Ch5.6).
- Planning for all combinations of N failure modes needs 2^N scenarios: 64 for 6, 1,024 for 10 (Ch5.6).
- Three failure modes (60%, 20%, 10%) covered 90% of manufacturing disruptions (Ch5.6).
- Delivery robot: 30% blockage probability x 4 s delay = 1.2 s expected vs 0.5 s contingency planning, 0.7 s net gain (Ch5.6).
Classification
- Patterns
- Contingency planningConditional plansPareto selection of failure modes
- Quality attributes
- Performance efficiency (ISO/IEC 25010)Reliability (ISO/IEC 25010 | NIST AI RMF: valid and reliable)
- Risks mitigated
- Replanning pauses on predictable failuresCombinatorial explosion of contingency branches
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.