Title :
Online Convex Programming and regularization in adaptive control
Author :
Raginsky, Maxim ; Rakhlin, Alexander ; Yüksel, Serdar
Author_Institution :
Dept. of Electr. & Comput. Eng., Duke Univ., Durham, NC, USA
Abstract :
Online Convex Programming (OCP) is a recently developed model of sequential decision-making in the presence of time-varying uncertainty. In this framework, a decision-maker selects points in a convex feasible set to respond to a dynamically changing sequence of convex cost functions. A generic algorithm for OCP, often with provably optimal performance guarantees, is inspired by the Method of Mirror Descent (MD) developed by Nemirovski and Yudin in the 1970´s. This paper highlights OCP as a common theme in adaptive control, both in its classical variant based on parameter tuning and in a more modern supervisory approach. Specifically, we show that: (1) MD leads to a generalization of classical adaptive control schemes based on recursive parameter tuning; (2) A supervisory controller switching policy that uses OCP to estimate system parameters from a sequence of appropriately regularized output prediction errors can flexibly adapt to presence or absence of output disturbances in the system.
Keywords :
adaptive control; convex programming; decision making; time-varying systems; uncertain systems; adaptive control; convex cost function; generic algorithm; mirror descent method; online convex programming; recursive parameter tuning; sequential decision-making; supervisory approach; supervisory controller switching policy; time-varying uncertainty; Adaptation model; Adaptive control; Delta modulation; Mirrors; Programming; Switches;
Conference_Titel :
Decision and Control (CDC), 2010 49th IEEE Conference on
Conference_Location :
Atlanta, GA
Print_ISBN :
978-1-4244-7745-6
DOI :
10.1109/CDC.2010.5717262