Accelerating model predictive control by online constraint removal

We propose a method for the acceleration of the online linear model predictive control (MPC) calculations with partial information on the explicit solution. We highlight two properties of the proposed approach: (i) It does not require to calculate the explicit solution first, and its computational effort grows only polynomially in the number of the constraints of the problem. The proposed approach can therefore be applied to problems that are too large for today´s explicit MPC methods. (ii) The method is not based on a specific type or implementation of the optimization algorithm and can therefore easily be combined with a variety of existing MPC implementations. The proposed approach is, to the knowledge of the authors, one of yet a few attempts to use the insight into the structure of the explicit MPC law in online MPC.