The Power of Reformulation: PDDL+ Validation Through Planning

PDDL+ allows the formal specification of systems representing mixed discrete-continuous representation, under both a discrete and a continuous dynamics; this expressiveness is pivotal in real-world applications. An important aspect is the capability of validating plans obtained by planning systems, and assessing their compliance against the domain’s model. Unfortunately, a very limited number of validation tools are capable of dealing with PDDL+ tasks. To overcome this problem, in this work we propose an approach that allows to exploit any domain-independent PDDL+ or PDDL2.1 planning engine for validating PDDL+ plans. We introduce a set of translations that, given a PDDL+ plan and the corresponding PDDL+ task, generate a new PDDL+ or PDDL2.1 whose solvability is bound to the validity of the considered plan. We empirically evaluate the usefulness of the proposed approach on a range of PDDL+ benchmarks under an interpretation of time that can be either continuous (through a PDDL+ translation) and discrete (through a PDDL+ or a PDDL2.1 translation).