Unconstrained networked decentralized model predictive control

Complex processes are naturally suitable to be controlled in a decentralized framework: centralized control solutions are often unfeasible in dealing with large scale plants and they are computationally prohibitive when the processes are too fast for the existing computational resources. In these cases, the resulting control problem is usually split into many smaller subproblems and the global requirements are guaranteed by means of a proper coordination. A coordination strategy based on a networked decentralized Model Predictive Control is proposed in this paper for improving the global control performances. The innovative solution is based on independent agents and on a local area network used for exchanging a reduced set of information. The proposed architecture guarantees satisfactory performance under strong interactions among subsystems. A stability analysis is presented for the unconstrained decentralized case and the provided stability results are employed for tuning the decentralized controller. Numerical simulations are given for testing and validating the proposed technique.