Count Downward

Numeric planning with abstraction heuristics, participant of the International Planning Competition 2026.

Count Downward is a numeric planner that extends Numeric Fast Downward with two families of numeric abstraction heuristics: pattern databases (Gnad et al., 2025; Fritzsche et al., 2026) and domain abstractions (Fritzsche et al., 2026). Several planners built on this codebase participated in the International Planning Competition 2026, in the optimal, satisficing, and agile tracks. The sequential portfolios that combine these configurations and adds the numeric landmark-cut heuristic are called Count Downward Together.

The implementation is available on GitHub. The repository receives only rudimentary maintenance; we recommend switching to PlanForge, a reimplementation in Rust.

References

2026

  1. Managing Infinite Abstractions in Numeric Pattern Database Heuristics
    Markus Fritzsche, Daniel Gnad, Mikhail Gruntov, and 1 more author
    In Proceedings of the Fortieth AAAI Conference on Artificial Intelligence (AAAI 2026), 2026
    PDF Poster Slides
  2. Domain-Abstraction Heuristics for Simple Numeric Planning
    Markus Fritzsche, Mikhail Gruntov, Alexander Shleyfman, and 1 more author
    In Proceedings of the 19th International Symposium on Combinatorial Search (SoCS 2026), 2026
    PDF

2025

  1. PDBs Go Numeric: Pattern-Database Heuristics for Simple Numeric Planning
    Daniel Gnad, Lee-or Alon, Eyal Weiss, and 1 more author
    In Proceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligence (AAAI 2025), 2025
    PDF