Learning hierarchical task networks for nondeterministic planning domains

This paper describes how to learn Hierarchical Task Networks (HTNs) in nondeterministic planning domains, where actions may have multiple possible outcomes. We discuss several desired properties that guarantee that the resulting HTNs will correctly handle the nondeterminism in the domain. We developed a new learning algorithm, called HTN-MAKERND, that exploits these properties. We implemented HTN-MAKERND in the recently-proposed HTN-MAKER system, a goal-regression based HTN learning approach. In our theoretical study, we show that HTN-MAKERND soundly produces HTN planning knowledge in low-order polynomial times, despite the nondeterminism. In our experiments with two nondeterministic planning domains, ND-SHOP2, a well-known HTN planning algorithm for nondeterministic domains, significantly outperformed (in some cases, by about 3 orders of magnitude) the well-known planner MBP using the learned HTNs.