Cooperative adaptive optimization for a class of nonlinear systems

In this paper, we propose an adaptive optimization technique to solve large scale optimization problems for nonlinear systems composed of multiple interacting agents. The framework is based on the concept of cooperative optimization where each agent addresses a specific local optimization problem that requires information from some or all other agents. An inexact penalty function method is used to solve the overall cooperative optimization problem. The closed-loop follows an adaptive gradient flow that leads to the tracking of local critical points with guaranteed transient performance.