Abstract
A representation of the adjoint process, which appears in a general version of the maximum principle for control systems described by Girsanov solutions of stochastic differential equations, is given in terms of the linearization of the state equation. The result is only valid when the optimal control and the coefficiencies in the state equation are smooth; however two examples show that the result can nevertheless be applied to the nonsmooth case, solving in particular the linear regulator and the “predicted miss” problems.

This publication has 7 references indexed in Scilit: