A Collective Neurodynamic System for Distributed Optimization with Applications in Model Predictive Control

A collective neurodynamic system is presented to distributed convex optimization subject to linear equality and box constraints in framework of an autonomous multiagent network. The overall objective to be minimized takes an additive form of multiple local objective functions. Agents in the system, each of which is modeled by a recurrent neural network, cooperatively, and autonomously develop their dynamic behaviors based on realtime interactions with their neighbors. In specific, each recurrent neural network has knowledge on a local objective only with no access to the overall objective function. It exchanges neuronal state with neighboring agents via a predefined communication topology. Guided by their individual neurodynamics and the collective efforts, all agents reach consensus at the global optimal solution to the distributed convex optimization. Illustrative examples are provided to substantiate the theoretical properties. A case study on model predictive control is reported to further validate the collective neurodynamic system.