Abstract
We study the first-order changes in the minimum value of a stochastic Bolza problem under nonlinear perturbations of its dynamics, objectives and constraints. The perturbations are lumped together into a finite-dimensional parameter x, which gives rise to a minumum-value function V(α). The techniques of nonsmooth analysis facilitate the computation of the generalized gradient of V at O, a set which relates the sensitivity of the problem to the extremals of the stochastic maximum principle. Our method of proof includes a new derivation of the stochastic maximum principle for problems with equality and inequality constraints, while the formula we give for ∂V(0) has important implications regarding the stability of the constraints and the interpretation of the multipliers.

This publication has 14 references indexed in Scilit: