Splitting Multiple Situations in Conditional Planning

Most current conditional planners are based on explicit representation of alternative outcomes of conditional actions. The explicit representation of possible outcomes induces a split in the search space which rapidly exponentially grows. A general strategy of minimal splitting of multiple situationl, for conditional planners, is introduced in this paper, in order to reduce the amount of computational resources required to manage conditional branches. The strategy is based on the definition of conditional formulas, they allow to represent set of situations avoiding explicit combinatorial representation of them. Conditional formulas are used in order to keep the search space and plan representation as small as possible while splitting on conditional branches is avoided if not pertinent to the problem at hand. The proposed strategy also minimises the use of observing and sensing operations during plan execution monitoring.