Diffusing-Horizon Model Predictive Control

We analyze a time-coarsening strategy for model predictive control (MPC) that we call diffusing-horizon MPC. This seeks to overcome the computational challenges associated with optimal problems span multiple timescales. The coarsening approach uses time discretization grid becomes exponentially more sparse as one moves forward in time. design is motivated by recently established property of known exponential decay sensitivity (EDS). states impact parametric perturbation at future decays backward establish conditions under which this holds constrained MPC formulation linear dynamics and costs. Moreover, show proposed scheme can be cast problem and, thus, condition holds. use HVAC plant case study real data demonstrate approach. Specifically, times reduced two orders magnitude while increasing closed-loop cost only 3%.