The exact penalties technique is applied to the problem of optimal control of a system described by ordinary differential equations. The obtained functional is substantially non-smooth but differentiable. The differential equations are considered as limitations and 'eliminated' by the penalty function introduction.