Time-distributed optimization for real-time model predictive control: stability, robustness, and constraint satisfaction (Q2184518)

First, it is presented a general system theoretic framework for analyzing a broad class of time-distributed optimization algorithms. This framework applies to any model predictive control feedback law that is locally input-to-state stable combined with any optimization algorithm whose convergence rate that is at least locally \(q\)-linear. Moreover, the existence of a joint region of attraction for the state and solution estimate is established. Next, the proposed theoretical framework is applied to a real-time iteration scheme which uses a Gauss-Newton hessian approximation and solves a single quadratic program per sampling instant. The robustness of this scheme is established. Also, the effect of the number of the sequential quadratic programming iterations performed at each sampling instant is analyzed and sufficient conditions for robust constraint satisfaction are obtained. At last, simulation results concerning a bicycle model of a sedan illustrate the proposed approach.