A Sound (But Incomplete) Polynomial Translation from Discretised PDDL+ to Numeric Planning

pddl+ is an expressive planning formalism that enables the modelling of domains having both discrete and continuous dynamics. Recently, two mappings for translating discretised pddl+ problems into a numeric a-temporal task have been proposed. Such translations produce a task of exponential or polynomial size w.r.t. the size of the native task. In this work, starting from the above-mentioned polynomial translation, we introduce a sound but not generally complete variant that has the potential to improve the performance of numeric planning engines. We define the subclass of problems where the variant is safely applicable, and we assess the advantages of such a translation.