Computational issues in model predictive control

From MaRDI portal





Model Predictive Control (MPC) is an effective but computationally demanding control design methodology. This paper surveys different techniques used to alleviate the on-line computational burden in MPC, and it concludes that research efforts towards implementation of MPC have been essentially in two directions: (1) to develop fast optimization solvers tailored to the needs of MPC and/or (2) to move the MPC computations off-line. In particular, the second approach can be ultimately pushed towards the explicit or approximate off-line determination of the MPC feedback control law; this can be indeed a viable and efficient solution for small problems where the complexity of the MPC feedback is manageable, but may easily suffer from the curse of dimensionality for problems of larger size, for which fast solvers provide therefore a valuable tool. A suitable combination between on-line computation, using fast solvers, and off-line computation should give the best option in many circumstances. For the entire collection see [Zbl 0969.00072].











This page was built for publication: Computational issues in model predictive control

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4546618)