Closed-loop predictions in model based predictive control of linear and nonlinear systems (Q2776232)

The closed-loop prediction paradigm is under investigation in order to address the invariance/feasibility problem of model predictive control (MPC). The strength of closed-loop prediction in handling MPC of linear uncertain systems and nonlinear systems is exposed. The overall MPC strategy is decomposed into the design of a fixed-term linear inner loop with a variable nonlinear loop wrapped around it. It is shown that the closed-loop prediction paradigm reduces on-line optimization to a feasibility problem, which can be solved off-line producing significant simplification. The proposed algorithms guarantee closed-loop stability and can asymptotically yield optimal dynamic behaviour. Two numerical examples illustrate the application and features of the closed-loop prediction approach.NEWLINENEWLINEFor the entire collection see [Zbl 0976.00019].