Single Network Adaptive Critic aided nonlinear dynamic inversion

Abstract

A Single Network Adaptive Critic (SNAC) [1] architecture has been shown to have much faster convergence than Adaptive Critic (AC) and offers an optimal control design method for state regulation. Dynamic Inversion (DI) [2], on the other hand, is a very popular nonlinear control design method for both output regulation and command tracking that offers a closed form expression for control but lacks the interpretation of optimal control design. Here, we propose a systematic technique where a pre-synthesized SNAC network aids the DI control so that the resulting control can be used for both near optimal regulation and command following. The potential of this new methodology is demonstrated by considering a benchmark nonlinear system.