Perturbation methods are used to implement the conditions of nonlocal improvement of controls, constructed in the form of special boundary value problems in space of phase and conjugate variables and in the form of special tasks of a fixed point of definite operator in the space of controls. Terms improvement of controls are determined by the operation of projection onto the set of admissible control values. Methods are characterized by a lack of operations of convex or needle variation of controls and principal possibility to improve suboptimal controls satisfying the maximum principle.