Planning with Primary E ects: Experiments and Analysis
نویسندگان
چکیده
The use of primary eeects in planning is an eeective approach to reducing search. The underlying idea of this approach is to select certain \important" eeects among the eeects of each operator and to use an operator only for achieving its important eeects. In the past, there has been little analysis of planning with primary eeects and few experimental results. We provide empirical and analytical results on the use of primary eeects. First, we experimentally demonstrate that the use of primary eeects may lead to an exponential reduction of the planning time. Second, we analytically explain the experimental results and identify the factors that innuence the eeciency of planning with primary eeects. Third, we describe an application of our analysis to predicting the performance of a planner for a given selection of primary eeects.
منابع مشابه
Automatically Selecting and Using Primary E ects in Planning: Theory and Experiments
Using primary e ects of operators in planning is an e ective approach to reducing planning time and improving solution quality. However, the characterization of \good" primary e ects has remained at an informal level. In addition, no method has previously been known to automatically learn the primary e ects of operators from a given domain speci cation. In this paper we formalize the use of pri...
متن کاملSearching for Planning Operators with Context-Dependent and Probabilistic E ects
Providing a complete and accurate domain model for an agent situated in a complex environment can be an extremely di cult task. Actions may have di erent e ects depending on the context in which they are taken, and actions may or may not induce their intended e ects, with the probability of success again depending on context. We present an algorithm for automatically learning planning operators...
متن کاملEecient Planning in Stochastic Domains through Exploiting Problem Characteristics
Partially observable Markov decision process (POMDP) can be used as a model for planning in stochastic domains. However, general POMDPs are computationally expensive to solve. This paper investigates how might problem characteristics be exploited to cut down computation. We consider planning problems where observations are informative of the world state and there are not much uncertainties in e...
متن کاملVery long baseline neutrino oscillation experiments and the MSW e!ectq
Assuming three-neutrino mixing, we study the capabilities of very long baseline neutrino oscillation experiments to verify and test the MSW e!ect and to measure the lepton mixing angle h 13 . We suppose that intense neutrino and antineutrino beams will become available in the so-called neutrino factories. We "nd that the most promising and statistically signi"cant results can be obtained by stu...
متن کاملAttributable E¤ects in Randomized Experiments and Observational Studies
The motivation for randomization inference and sensitivity analysis is reviewed. New methods are discussed for inverting randomization tests to provide interval estimates of the magnitude of treatment e¤ects. 1 Notation and Review
متن کامل