Linear Tracking MPC for Nonlinear Systems—Part II: The Data-Driven Case

We present a novel data-driven model predictive control (MPC) approach to unknown nonlinear systems using only measured input-output data with closed-loop stability guarantees. Our scheme relies on the system parametrization provided by Fundamental Lemma of Willems et al. use new measurements online update data, exploiting local linear approximations underlying system. prove that our MPC scheme, which requires solving strictly convex quadratic programs online, ensures closed loop (practically) converges (unknown) optimal reachable equilibrium tracks desired output reference while satisfying polytopic input constraints. As intermediate results independent interest, we extend affine and derive robustness bounds w.r.t. noisy for open-loop problem, are directly transferable other schemes in literature. The applicability is illustrated numerical application continuous stirred tank reactor.