The paper addresses the problem of numerical issues and degeneracies arising when solving the parametric linear complementarity problem (pLCP) for the purpose of computing partitions of explicit model predictive controllers (eMPC) based on constrained linear models and a 2-norm cost function. After summarizing the basic pLCP algorithm, several numerical issues relevant for reliable computation of eMPC partitions are exposed. The improved performance is illustrated with an eMPC controller example, which poses a problem for the available multi-parametric quadratic programming (pQP) solvers.
COBISS.SI-ID: 24318503
The paper addresses numerical issues and degenerations in the parametric quadratic programming (pQP) algorithm, used for computing partitions of explicit predictive controllers in the Multi-Parametric Toolbox. We summarise the pQP problem and the basic algorithm, analyse its implementation in the MPT toolbox, expose the numerical issues and suggest a series of improvements for more reliable operation.
COBISS.SI-ID: 24307239