Time-Varying Optimization of Networked Systems With Human Preferences

This article considers a time-varying optimization problem associated with network of systems, each the systems shared by (and affecting) number individuals. The objective is to minimize cost functions individuals' preferences, which are unknown, subject constraints that capture physical or operational limits network. To this end, develops distributed online algorithm concurrent learning functions. learned on-the-fly based on users' feedback (provided at irregular intervals) leveraging tools from shape-constrained Gaussian process. primal–dual method and acts effectively in closed-loop fashion where: First, utilized estimate cost, second, measurements algorithmic steps bypass need for sensing (unknown) exogenous inputs performance analyzed terms dynamic regret constraint violation. Numerical examples presented context real-time energy resources.