An algorithm for multi-parametric quadratic programming and explicit MPC solutions (Q1868059)

A new multi-parametric quadratic program (mp-QP) solver for constrained model predictive control (MPC) is proposed. The properties of the polyhedral partition of the state space induced by the multi-parametric piecewise affine (PWA) solutions are studied, and the explicit solution for linear constrained MPC is obtained. Based on the exploitation of direct relations between neighboring polyhedral regions and combinations of active constraints, a new exploration strategy for subdividing the parameter space is derived. The strategy avoids: an unnecessary partition, the solution of linear programs for determining an interior point in each new region of the parameter space, and the solution of the QP problem for the mentioned interior points. Numerical examples indicate large improvements of computational efficiency over existing mp-QP algorithms.