Automated Planning and Acting

ion Heuristics An abstraction of a planning domain Σ is a γ-preserving homomorphism from Σ onto a smaller planning domain Σ′. For each planning problem P = (Σ, s0, g), this defines a corresponding abstraction P ′ = (Σ′, s0, g ′); and if we let c∗ denote the cost of an optimal solution to a planning problem, then it follows that c∗(P ′) ≤ c∗(P ). If Σ′ is simple enough that we can compute c∗(P ′) for every planning problem P ′ in Σ′, then the function h(s) = c∗(Σ′, s′, g′) is an admissible heuristic for P . Authors’ manuscript. Published by Cambridge University Press. Do not distribute.