State-flipped control and Q-learning for finite horizon output tracking of Boolean control networks

Abstract

This article explores the state-flipped control mechanism for the finite horizon output tracking of Boolean control networks (BCNs) subject to a time-varying reference output trajectory. Firstly, the concept of joint control pair consisting of state-flipped control and traditional control is proposed, based on which, the joint reachability set of BCNs with joint control pair is given. Secondly, combining the joint reachability set with the output-correlation set, a necessary and sufficient condition is presented to determine the solvability of the finite horizon output tracking problem. Thirdly, a Q-learning based algorithm is developed to find both the minimum combination flip set and joint control pair sequence. Finally, an example is given to illustrate the proposed theoretical results.