Abstract. An efficient Newton-type scheme for the approximate on-line solution of optimization problems as they occur in optimal feedback control is presented. The scheme allows a fast reaction to disturbances by delivering approximations of the exact optimal feedback control which are iteratively refined during the runtime of the controlled process. The contractivity of this real-time iteration scheme is proven, and a bound on the loss of optimality-compared with the theoretical optimal solution-is given. The robustness and excellent real-time performance of the method is demonstrated in a numerical experiment, the control of an unstable system, namely, an airborne kite that shall fly loops. 1. Introduction. Feedback control based on the real-time optimization of nonlinear dynamic process models, also referred to as nonlinear model predictive control (NMPC), has attracted increasing attention over the past decade, particularly in chemical engineering [4,27,1,28]. Based on the current system state, feedback is provided by an online optimization of the predicted system behavior, using the mathematical model. The first part of the optimized control trajectory is implemented at the real system, and a sampling time later the optimization procedure is repeated. Among the advantages of this approach are the flexibility provided in formulating the objective and in modeling the process using ordinary or partial differential equations (ODEs or PDEs), the capability of directly handling equality and inequality constraints, and the possibility of treating large disturbances quickly.One important precondition, however, is the availability of reliable and efficient numerical optimal control algorithms. One particularly successful algorithm that is designed to achieve this aim, the recently developed real-time iteration scheme, will be the focus of this paper. In the literature, several suggestions have been made on how to adapt off-line optimal control algorithms for use in on-line optimization. For an overview and comparison of important approaches, see, e.g., Binder et al. [6]. We particularly mention here the "Newton-type control algorithm" proposed by Li and Biegler [32] and de Oliveira and Biegler [15] and the "feasibility-perturbed SQP" approach to NMPC by Tenny, Wright, and Rawlings [38]. Both approaches keep even intermediate optimization iterates feasible. This is in contrast to the simultaneous dynamic optimization methods, as the collocation method proposed in Biegler [5] or the direct multiple shooting method in Bock et al. [7] and in Santos [37], which allow