Abstract
This article deals with optimizing problems whose restrictions are nonholonomic. The central issue relates to dual nonholonomic programs (what they mean and how they are solved?) when the nonholonomic constraints are given by Pfaff equations. We emphasize that nonholonomic critical points are not the classical ones and that the nonholonomic Lagrange multipliers are not the classical (holonomic) Lagrange multipliers. Topological significance of Lagrange multipliers and dual function theory introduced by EDO and EDP are key results. Also new Riemannian geometries attached to a given nonholonomic constrained optimization problem are introduced. The original results are surprising and include: (i) aspects derived from the Vranceanu theory of nonholonomic manifolds, and from the geometric distributions theory, (ii) optimal problems in Darboux canonical coordinates.