One of the major techniques to tackle temporal planning problems is heuristic search augmented with a symbolic representation of time in the states. Augmenting the problem with composite actions (macro-actions) is a simple and powerful approach to create "shortcuts" in the search space, at the cost of augmenting the branching factor of the problem and thus the expansion time of a heuristic search planner. Hence, it is of paramount importance to select the right macro-actions and minimize the number of such actions to optimize the planner performance. In this paper, we first discuss a simple, yet powerful, model similar to macro-actions for the case of temporal planning, and we call these macro-events. Then, we present a novel ranking function to extract and select a suitable set of macro-events from a dataset of valid plans. In our ranking approach, we consider an estimation of the hypothetical search space for a blind search including a candidate set of macro-events under four different exploitation schemata. Finally, we experimentally demonstrate that the proposed approach yields a substantial performance improvement for a state-of-the-art temporal planner.

Learning domain knowledge from small training problems to improve planning performance on arbitrarily sized problems is a highly active research area. Many works explored the use of macro-actions to create “shortcuts” in the search space, at the cost of increasing the branching factor of the problem. In temporal planning, a recent technique proposes to equip a heuristic-search temporal planner with selected “macro-events”: a “shortcut” mechanism similar to macro-actions but with state-dependent semantics. In this paper, we generalize macro-events to a lifted representation, making them independent of specific problem objects. We devise a fully automated framework that, given a domain and a collection of small training problems, constructs and selects a suitable set of lifted macro-events. We define a learning pipeline that mixes the optimization of the statistical expectation on an abstraction of the problem with an empirical refinement of the selection on a validation set. We experimentally show that the proposed approach scales to complex problems, yielding substantial improvements over the baseline.