Combining The Delete Relaxation With Critical-Path Heuristics: A Direct Characterization

Recent work has shown how to improve delete relaxation heuristics by computing relaxed plans, i.e., the h(FF) heuristic, in a compiled planning task Pi(C) which represents a given set C of fact conjunctions explicitly. While this compilation view of such partial delete relaxation is simple and elegant, its meaning with respect to the original planning task is opaque, and the size of Pi(C) grows exponentially in vertical bar C vertical bar. We herein provide a direct characterization, without compilation, making explicit how the approach arises from a combination of the delete-relaxation with critical-path heuristics. Designing equations characterizing a novel view on h(+) on the one hand, and a generalized version h(C) of h(m) on the other hand, we show that h(+) (Pi(C)) can be characterized in terms of a combined h(C+) equation. This naturally generalizes the standard delete-relaxation framework: understanding that framework as a relaxation over singleton facts as atomic subgoals, one can refine the relaxation by using the conjunctions C as atomic subgoals instead. Thanks to this explicit view, we identify the precise source of complexity in h(FF) (Pi(C)), namely maximization of sets of supported atomic subgoals during relaxed plan extraction, which is easy for singleton-fact subgoals but is NP-complete in the general case. Approximating that problem greedily, we obtain a polynomial-time h(CFF) version of h(FF) (Pi(C)), superseding the Pi(C) compilation, and superseding the modified Pi(C)(ce) compilation which achieves the same complexity reduction but at an information loss. Experiments on IPC benchmarks show that these theoretical advantages can translate into empirical ones.