Data-driven Tube-Based Stochastic Predictive Control

A powerful result from behavioral systems theory known as the fundamental lemma allows for predictive control akin to Model Predictive Control (MPC) linear time-invariant (LTI) with unknown dynamics purely data. While most data-driven literature focuses on robustness respect measurement noise, only a few works consider exploiting probabilistic information of disturbances performance-oriented in stochastic MPC. This work proposes novel scheme chance-constrained LTI subject noise and additive disturbances. In order render otherwise intractable optimal problem deterministic, our approach leverages ideas tube-based MPC by decomposing state into deterministic nominal driven inputs error affected Satisfaction original chance constraints is guaranteed tightening probabilistically robustly noise. The resulting receding horizon lightweight, recursively feasible, renders closed loop input-to-state stable presence both We demonstrate effectiveness proposed simulation example.