Heuristics for Bounded-Cost Search

Abstract

The problem of searching for a plan with cost at most equal to a given absolute bound has attracted interest recently, and several search algorithms tailored specifically to this problem have been proposed. We investigate instead how to adapt planning heuristics to this setting. A few of the resulting heuristics, used in a greedy bounded-cost search, perform better than previous and baseline methods, but only by a small margin. Making effective use of the cost bound in bounded-cost planning, it appears, remains a challenge.