Existence of an optimal control with sparse jumps in the state variable