Translations from Discretised PDDL+ to Numeric Planning

Abstract

Hybrid PDDL+ models are amongst the most advanced models of systems and the resulting problems are notoriously difficult for planning engines to cope with. An additional limiting factor for the exploitation of PDDL+ approaches in real-world applications is the restricted number of domain-independent planning engines that can reason upon those models. With the aim of deepening the understanding of PDDL+ models, in this work we study a novel mapping between a time discretisation of PDDL+ and numeric planning as for PDDL2.1 (level 2). The proposed mapping not only clarifies the relationship between these two formalisms, but also enables the use of a wider pool of engines, thus fostering the use of hybrid planning in real-world applications. Our experimental analysis shows the usefulness of the proposed translation, and demonstrates the potential of the approach for improving the solvability of complex PDDL+ instances.

The Power of Reformulation: From Validation to Planning in PDDL+

“…As shown by Percassi, Scala, and Vallati (2021), a PDDL+ problem can be transformed into a numeric one using two different translations, i.e., EXP and POLY. Such encodings called Π EXP and Π POLY can be used to seek discrete valid PDDL+ plans.…”

“…The use of engines that exploit a discretisation-based approach to PDDL+, such as ENHSP (Scala et al 2016), allows demonstrating the validity of a plan with respect to the discrete semantics, while the use of engines capable of reasoning over a continuous timeline, like SMTPLAN (Cashmore, Magazzeni, and Zehtabi 2020), allows to test the validity with regards to the continuous semantics. Further, leveraging the work by Percassi, Scala, and Vallati (2021) to translate PDDL+ tasks into numeric PDDL2.1 (level 2) tasks, we introduce an optimised translation that allows extending the pool of planning engines to include those that support PDDL2.1. Our approach enables the validation of domain models using non-polynomial dynamics, and the validation of plans under discrete semantics (Percassi, Scala, and Vallati 2021).…”

“…Further, leveraging the work by Percassi, Scala, and Vallati (2021) to translate PDDL+ tasks into numeric PDDL2.1 (level 2) tasks, we introduce an optimised translation that allows extending the pool of planning engines to include those that support PDDL2.1. Our approach enables the validation of domain models using non-polynomial dynamics, and the validation of plans under discrete semantics (Percassi, Scala, and Vallati 2021). We are not aware of approaches that are capable of handling these two aspects.…”

“…Following Percassi, Scala, and Vallati (2021) and Shin and Davis (2005), we formalise the PDDL+ semantics through the notion of time points, histories and the projection of a plan given a domain. We assume the reader is familiar with well known notions of action/event applicability, and use in the rest γ(s, •) to denote the state resulting by applying either an action/event (γ(s, a)) or a sequence of action/events (γ(s, a 1 , ..., a n )) in a state s.…”

LM-Cut Heuristics for Optimal Linear Numeric Planning

“…ROVER-METRIC and ZENOTRAVEL-LINEAR are introduced by Leofante et al (2020) and also have no SOSE. The domains prefixed by LIN-CAR-are compiled from PDDL+ domain LIN-CAR (Fox and Long 2006) using a recently proposed method (Percassi, Scala, and Vallati 2021), the details of which are described in the SM. Some linear effects in LIN-CAR-EXP are SOSE, but cost(â u ) = 0 for each âu , a ∈ supp 2 (ψ), which results in m v âu,a (s, ψ) = 1 and does not make much difference between h LM-cut 1 and h LM-cut 2 .…”