In this paper, an extended approach to the nonlinear DMC algorithm is proposed, which can handle constrained and MIMO systems under some defined conditions. This extension also admits higher control horizon, M and prediction horizon, P that are required to increase the performance of the controller in the most practical applications. Simulation results of the proposed method in the control of a stirred tank reactor with 2 inputs, 2 outputs nonlinear model and also a power unit nonlinear model with 3 inputs, 3 outputs are presented to illustrate effectiveness of the method and its justification. (c) 2005 Elsevier Ltd. All rights reserved.