Representing Pattern Databases with Succinct Data Structures

摘要

In this paper we describe novel representations for pre-computed heuristics based on Level-Ordered Edge Sequence (LOES) encodings. We introduce compressed LOES, an extension to LOES that enables more aggressive compression of the state-set representation. We evaluate the novel representations against the respective perfect-hash and binary decision diagram (BDD) representations of pattern databases in a variety of STRIPS domains.