Faster planning algorithms for AI by fixing a math mismatch

Researchers found that a widely-used algorithm for automated planning (helping AI systems find solutions to problems) was using a math formula designed for a different situation, which made it slower and less reliable. Fixing this mismatch — by accounting for the fact that different problems have different-sized rewards — makes planning algorithms find good solutions faster and with fewer computational steps.

This is an incremental improvement to how planning algorithms work internally; it matters mostly to researchers building better AI planning systems, not to people using those systems.