A Discussion of the Relationships Between the Dual Goals of Stochastic Control.

摘要

The possibility of aligning the dual goals of an optimal stochastic controller is discussed. It is suggested that when the measurement function is chosen so that these two dual goals are aligned an artificial separation will occur. This will occur since the action taken to follow a trajectory will also lead to the best possible use of the control for estimation purposes. A simple set of examples describing the nature of such alignment and cases of nonalignment is given. (Author)