Universal planning: an (almost) universally bad idea

Several authors have recently suggested that a possible approach to planning in uncertain domains is to analyze all possible situations beforehand, and to then store information about what to do in each. The result of this is that a system can simply use its sensors to examine its domain, and then decide what to do by finding its current situation in some sort of a table. The purpose of this note is to argue that even if the compile-time costs of the analysis are ignored, the size of the table must in general grow exponentially with the complexity of the domain. This makes it unlikely that this approach will be able to deal with problems of an interesting size; one really needs the ability to do some amount of inference at run-time. Alternatively put, an effective approach to acting in uncertain domains cannot be to look and then leap, but must always be to look, to think, and only then to leap. 1 There are too many universal plans In order for us to present a sharp criticism of the a...