Closed-loop stochastic dynamic process optimization under input and state constraints

We introduce a concept to solve closed-loop dynamic optimization problems. The key aspect in our approach is the simultaneous optimization over a reference trajectory (steering task) and a feedback map (controlling task) while dynamically sizing back-off to the state and inputs constraints. This approach to constrained dynamic optimization and control distinguishes itself from any existing open-loop strategy (including model predictive control) by explicitly predicting the closed-loop behavior of the plant under the influence of stochastic disturbances. We show by employment of a proper process operation framework that this is a convex conic optimization problem for which efficient interior-point algorithms are available making the problem numerically tractable.