Mixed Propositional and Numeric Planning in the Model Checking Integrated Planning System

The paper considers the extensions to the domainindependent Model Checking Integrated Planning System (MIPS) to pre-processes and solve mixed propositional and numerical planning problems in PDDL+ syntax for the 3rd international planning competition. The static analyzer grounds all predicates and functors, distinguishes constant from fluent atoms and numerical constants from variables. It further approximates the bounding intervals for the resource variables, and encodes their possible finite domain. Pre-compilation also establishes symmetries within the object set and dependencies among the set of operators. Based on the inferred information, the directed search exploration algorithm applies critical path scheduling to parallelize sequential plans and to refine a relaxed plan graph heuristic, while different pruning approaches effectively reduce the branching factor.