A novel robust predictive control algorithm for input-saturated uncertain linear discrete-time systems with structured norm-bounded uncertainties is presented. The solution is based on the minimization, at each time instant, of a LMI convex optimization problem obtained by a recursive use of the S-procedure. The general case of N free moves is presented. Stability and feasibility are proved and comparisons with robust multi-model (polytopic) MPC algorithms are also presented via an example.