Modeling Probabilistic Actions for Practical Decision-Theoretic Planning
AnHai Doan · 1996
Most existing decision-theoretic planners represent uncertainty in the state of the world with a precisely specified probability distribution over world states. This representation is not expressive enough to model many interesting classes of practical planning problems. and renders inapplicable some abstractionbased planning approaches. In this paper we propose as a remedy a more general action model with a well-founded semantics based on probability intervals. In particular, sets of probability distributions representing uncertainty about the world are represented with mass assignments, the representation of which is slightly modified to suit practical considerations. We present a projection rule and prove it correct. Complexity results and empirical evidence are also provided which suggest evaluating plans (projecting plans and computing the expected utility) in our framework is efficient, and the action model is applicable in real-world domains. Introduction Any planning model th...