Nonlinear MPC
MPC: Model Predictive Control
Overview
NMPC plans a sequence of control actions by predicting how a nonlinear system will evolve, applying the first action, and replanning as new measurements arrive. Common approaches solve the resulting optimization with methods such as sequential quadratic or convex programming, while sampling-based methods explore candidate action sequences; learned models and reduced-order dynamics can make prediction more practical. Safety and robustness are addressed with tools including control barrier functions, reachability bounds, uncertainty-aware constraint tightening, and state estimation. A central challenge is achieving fast, feasible optimization with useful prediction horizons while retaining reliable constraint satisfaction under model error, disturbances, and changing environments.
Latest papers
82 papers12 papers in the last four weeks, up 140% on the four weeks before. 0.1% of all new papers.