A Real-Time Trajectory-Based Stability Constraint for Model Predictive Control

A new real time stability constraint for model predictive control is developed in this paper. Motivated by the dissipative system theory, the proposed stability condition is based on the controller trajectory and can be implemented on-the-fly to ensure the stability of MPC systems. The stability bound on the immediate-future values of manipulated variables is iteratively computed from the resultant stability condition. It is then used as a dynamic stability constraint in a typical MPC algorithm. The preliminary simulation studies with stable linear plants have shown a successful stabilization of very short predictive horizons.