ِContribution à l'étude des contröles optimales stochastiques

Abstract

In this work, we focus on the necessary conditions of optimality in stochastic optimal control the system is governed by an EDS. These necessary conditions are established in the form of maximum principle and demonstrated two new results: The first result concerns the maximum principle for singular diffusions which are non-linear coefficient; again, we do not assume that the coefficients of the cost function are convex. The result will be obtained using the weak perturbation of controls and a simple variational method. This is a generalization of the result obtained by Cadellinas-Haussmann and as that obtained by Benoussan.