Internal Model-Based Online Optimization

In this paper we propose a model-based approach to the design of online optimization algorithms, with goal improving tracking solution trajectory (trajectories) w.r.t. state-of-the-art methods. We focus first on quadratic problems time-varying linear term, and use digital control tools (a robust internal model principle) novel algorithm that can achieve zero error by modeling cost dynamical system. prove convergence for both strongly convex problems. further discuss sensitivity proposed method uncertainties quantify its performance. how be applied general (non-quadratic) using an approximate cost, analyze leveraging small gain theorem. present numerical results showcase superior performance algorithms over previous methods non-quadratic