A patchy approximation of explicit model predictive control

Abstract : The explicit solution of multi-parametric optimisation problems (MPOP) has been used to construct an off-line solution to relatively small- and medium-sized constrained control problems. The control design principles are based on receding horizon optimisation and generally use linear prediction models for the system dynamics. In this context, it can be shown that the optimal control law is a piecewise linear (PWL) state feedback defined over polytopic cells of the state space. However, as the complexity of the related optimisation problems increases, the memory footprint and implementation of such explicit optimal solution may be burdensome for the available hardware, principally due to the high number of polytopic cells in the state-space partition. In this article we provide a solution to this problem by proposing a patchy PWL feedback control law, which intend to approximate the optimal control law. The construction is based on the linear interpolation of the exact solution at the vertices of a feasible set and the solution of an unconstrained linear quadratic regulator (LQR) problem. With a hybrid patchy control implementation, we show that closed-loop stability is preserved in the presence of additive measurement noise despite the existence of discontinuities at the switch between the overlapping regions in the state-space partition.
Document type :
Journal articles
Complete list of metadatas

https://hal-supelec.archives-ouvertes.fr/hal-00751235
Contributor : Josiane Dartron <>
Submitted on : Tuesday, November 13, 2012 - 9:03:17 AM
Last modification on : Tuesday, September 18, 2018 - 1:02:09 AM

Identifiers

Collections

Citation

Nam Nguyen, Sorin Olaru, Morten Hovd. A patchy approximation of explicit model predictive control. International Journal of Control, Taylor & Francis, 2012, 85 (12), pp.1929-1941. ⟨10.1080/00207179.2012.713516⟩. ⟨hal-00751235⟩

Share

Metrics

Record views

164