A method of designing a type-one optimal servo system is described. The design uses an exact model matching method with pole-zero cancellations. The initial values of the integrators are set according to the plant initial conditions at the beginning of control. This servo system is optimal in the sense that the performance index, comprising the quadratic forms of both the error vector and the deviation control vector from its steady-state, is minimized. The sensitivity to plant parameter variations can be reduced without loss of optimality for the nominal plant. Moreover, this servo system has the same robustness characteristic as typical optimal servo systems designed using different performance indices. Two numerical examples illustrate the efficiency of this servo system.