The paper concerns the issue of optimum control of the strongly non-linear dynamic system, i.e. Anti-Tank Guided Missile (ATGM). The linear-quadratic regulator (LQR) was used to provide control capabilities. In order to use the classic LQR, the dynamics of the object must be presented in the form of a linear-stationary model. This is not possible in the case of the considered missile, mostly due to mass changing in time (intensive consumption of fuel) and varying aerodynamic conditions depending on the Mach number Ma. Thus, we are dealing with a non-stationary system. Moreover, state variables are frequently involved in complex functions, which do not allow one to separate coefficients related to state variables very easily. In order to linearize such a complex system, the paper uses Jacobian, as the matrix of state, calculated at each time instant. The automatic pilot of the ATGM, using the LQR method, determines the signals controlling the angles of flight control surfaces and the thrust vector using continuously calculated Jacobians. The paper presents the algorithm for the ATGM control.