Learning Features and Abstract Actions for Computing Generalized Plans
By Blai Bonet, Guillem Francès, Hector Geffner. 2018
69 citations recorded by the source corpus.
Generalized planning is concerned with the computation of plans that solve not one but multiple instances of a planning domain. Recently, it has been shown that generalized plans can be expressed as mappings of feature values into actions, and that they can often be computed with fully observable non-deterministic (FOND) planners. The actions in such plans, however, are not the actions in the instances themselves, which are not necessarily common to other instances, but abstract actions that are defined on a set of common features. The formulation assumes that the features and the abstract actions are given. In this work, we address this limitation by showing how to learn them automatically. The resulting account of generalized planning combines learning and planning in a novel way: a learner, based on a Max SAT formulation, yields the features and abstract actions from sampled state transitions, and a FOND planner uses this information, suitably transformed, to produce the general plans. Correctness guarantees are given and experimental results on several domains are reported.
Subfields: Artificial Intelligence
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Hector Geffner
- Tomas Lozano-Perez
- Leslie Pack Kaelbling
- Josh Tenenbaum
- Blai Bonet
- Tom Silver
- Sergey Levine
- Chelsea Finn
Citations in the corpus
- Learning First-Order Symbolic Representations for Planning from the Structure of the State Space (2019, methodology): “The same holds for recent methods for learning abstract planning representations [7], while others methods just require states to be factored over a set of low-level variables [24].”
- Discovering State and Action Abstractions for Generalized Task and Motion Planning (2021, methodology): “We chose tasks that are physical instantiations of widely used tasks in generalized planning literature (Srivastava, Immerman, and Zilberstein 2011; Bonet, Francès, and Geffner 2018; Illanes and McIlraith 2019) or classically difficult TAMP domains with large numbers of objects.”