A First Theory of Plausible Inference and Its Use in Continuous Domain Planning

摘要: Publisher Summary This chapter discusses the first theory of plausible inference and its use in continuous domain planning. The primary interest lies explanation-based learning. Plausibility allows explanations to be conjectured rather than entailed by background knowledge. shift away from entailment absolute truth takes pressure off system's inferential ability, thereby allowing nonbrittle behavior. Plausible are constructed qualitative statements describing domain. These mediate between numerical values more conventional symbolic representations. control analyses that calibrated with observed expert's behavior refined if necessary when deficiencies course exercising planning system. Planning, apart execution, consists decisions about how manipulate controllable parameters. Planning time is sum CPU expended over all steps. includes computation quantitative concept refinement.