Aspects of Intuitive Control: Stabilize, Optimize, and Identify

Abstract

This paper presents a novel interpretation to intuitive control with aspects of adaptive as well as optimal control. The regulation of system states in an iterative fashion to approach an optimal controller is presented. The system model is assumed to be linear time invariant with unknown internal dynamics. The adaptive portion employs a Model Reference Adaptive controller to stabilize the plant and subsequently provide an initial stabilizing gain for a Reinforcement Learning based policy iteration algorithm. The objective based decomposition of the controller into stabilization and optimization phases can prove instrumental in applications of robotics and aircraft navigation. A novel method for identification of unknown internal dynamics (MIMO plant) is also presented based on information from Optimization phase. The performance of this approach is demonstrated on an aircraft lateral-directional control application.