Abstract

Multivariate optimal control schemes consider the interactions between elements composing the interactive system, where model Predictive Control-MPC-is one of its most successful approaches. However, as the number of states and interactions increase, they may become impractical for real-time applications. Distributed MPC strategies aim to tackle these issues by solving several local MPC problems. But most of these strategies force the subsystems to cooperate or still require several iterations per time sample that may increase the computational and communication burden. In this work, a novel DMPC based on bargaining games approach is presented. The aim of this strategy is to allow subsystems to cooperate only if they perceive any benefit of it. Additionally, the proposed negotiation model avoids iterative procedures at each time step, while sharing small amounts of information. A plant of two stirred tanks followed by a flash separator was used for testing the proposed approach, which was compared with a centralized and a fully cooperative MPC strategies. The results evidence the better behavior of the proposed approach despite its single iteration per time step, making it feasible to implement in low capability systems.