According to a new arXiv preprint, reduced-order models (ROMs) make model predictive control (MPC) of high-dimensional nonlinear systems tractable, but the modeling error they introduce can cause constraint violations on the real system. The authors propose a framework called CoRE-MPC to address this problem.
The abstract frames the issue as a trade-off: ROMs lower computational cost, yet the gap between the reduced model and the full-order system threatens safety. CoRE-MPC is described as a robust method, though the full technical details are not available in the abstract.
The work targets nonlinear systems, where full-order MPC is often impractical. By accounting for modeling error, CoRE-MPC aims to preserve the tractability of reduced-order control while maintaining safety guarantees on the original system.