12.2.3 Probabilistic Methods for Localization

The localization problems considered so far have involved only nondeterministic uncertainty. Furthermore, it was assumed that nature does not interfere with the state transition equation or the sensor mapping. If nature is involved in the sensor mapping, then future I-states are not predictable. For the active localization problem, this implies that a localization plan must use information feedback. In other words, the actions must be conditioned on I-states so that the appropriate decisions are taken after new observations are made. The passive localization problem involves computing probabilistic I-states from the sensing and action histories. The formulation and solution of localization problems that involve nature and nondeterministic uncertainty will be left to the reader. Only the probabilistic case will be covered here.