Input-to-State Stability of Newton Methods in Nash Equilibrium Problems with Applications to Game-Theoretic Model Predictive Control

원문 초록

We prove input-to-state stability (ISS) of perturbed Newton methods for generalized equations arising from Nash equilibrium (NE) and generalized NE (GNE) problems. This ISS property allows the use of inexact computation in equilibrium-seeking to enable fast solution tracking in dynamic systems and plays a critical role in the analysis of interconnected plant-optimizer feedback systems such as model predictive control (MPC) with iterative optimization algorithms. For NE problems, we address the local convergence of perturbed Newton methods from the variational inequality (VI) stability analysis and point out that the ISS of the perturbed Josephy-Newton can be established under less restrictive stability/regularity conditions compared to the existing results established for nonlinear optimizations. Agent-distributed Josephy-Newton methods are then developed to enable agent-wise local computations. For GNE problems, since they cannot be reduced to VI problems in general, we use semismooth Newton methods to solve the semismooth equations arising from the Karush-Kuhn-Tucker (KKT) systems of the GNE problem and show that the ISS of the perturbed semismooth Newton methods can be established under a quasi-regularity condition. Agent-distributed updates are also developed for the GNE-seeking. To illustrate the use of the ISS in dynamic systems, applications to constrained game-theoretic MPC (CG-MPC) are studied. A real-time CG-MPC solver with both time- and agent- distributed computations is developed and is proven to have bounded solution tracking errors.