An Efficient Model Predictive Control Approach via Parameterized Gaussian Kernels

This article proposes an efficient robust model predictive control (MPC) framework for constrained linear time-invariant (LTI) systems in the presence of additive disturbance. In the framework, a novel MPC optimization problem is formulated by using control parameterization based on Gaussian kernels. Then, an efficient MPC control law is derived, which aims at reducing the online computational cost. We show that the controlled system driven by the designed control law complies with system constraints. Moreover, the robust stability (i.e., the state trajectory converges into a bounded set) of the closed-loop system is obtained given feasibility is fulfilled and the Gaussian kernels are properly designed. Numerical examples and comparisons with the conventional robust MPC are performed, which validate the proposed method.