This paper proposes a new robust distributed model predictive control framework that uses a closed-loop dual-mode approach to reduce the demanding computations required to solve the on-line constrained optimization problem. The proposed algorithm requires solving N convex optimization problems in parallel based on exchange of information among the controllers. A relaxation technique is also developed to overcome the problem of feasibility for the initial iteration. Two simulation examples are used to illustrate the new method and for comparing the proposed algorithm with a previously developed technique in terms of performance and maximum CPU time per control interval. The simulation results showed that the new algorithm provides a significant reduction in online computations while resulting in comparative performance as compared to a previously reported algorithm. (C) 2012 Elsevier Ltd. All rights reserved.