Optimum impulsive midcourse guidance with control dependent errors