Model predictive control for non-linear missiles

The present paper reports an attempt of applying model predictive control (MPC) to design an autopilot for a non-linear missile. The non-linear, fast dynamics of the missile raise three issues in the design of an MPC algorithm: the choice of the MPC performance index, in particular the terminal weighting term, to compromise the performance and the stability requirements; loss of the global minimum in the online optimization since it is a non-linear optimization; and the computational time limitation imposed by the fast sampling requirement. For the first issue, a procedure is developed to determine the terminal weighting term using a new representation of the control sequence in the moving horizon. For the other two issues, a new initial control profile and an associated control strategy are adopted in each optimization routine. It is shown that the new MPC algorithm can guarantee stability, even when a local minimum is attained in the online optimization or the optimization process has to stop owing to the limitation of the sampling time. Simulation results carried on the missile show that good performance and stability are achieved by the new MPC algorithm, whereas four other current MPC algorithms lose their stability.