Comparative Analysis of Distributed Model Predictive Control Strategies

In this paper, a comparative analysis of two distributed model predictive control (DMPC) strategies tested in simulation on the same test bench in a multiple-agent framework is proposed. Firstly, some details about the non-cooperative DMPC with state-space `velocity-form' formulation called DMPC
<sub>ss</sub>
are given. After that, the non-cooperative DMPC with input-output model denoted DMPC
<sub>io</sub>
is briefly presented. Moreover, the distributed strategies are evaluated with respect to the centralized and decentralized model predictive control (MPC) strategies with `velocity-form' formulation, denoted C-MPC
_ss
and DC-MPC
_ss
, respectively. The simulation results obtained on the octuple tank process, defined as two quadruple tank processes coupled in series, show that the two non-cooperative algorithms perform similarly with the centralized and decentralized strategies when tuned properly.