Data-based control of Logical Networks
Giorgia Disarò, Maria Elena Valcher
TL;DR
This work develops a data-driven framework for Boolean Control Networks (BCNs) using informativity to solve three core control problems—state-feedback stabilization, safe control, and output regulation—without identifying the full network model. By treating all BCNs compatible with the collected data as candidates, it derives necessary and sufficient data conditions and provides constructive state-feedback laws that guarantee the target behavior for the true BCN. The approach builds on the algebraic BCN representation via semi-tensor products and develops algorithms to extract equilibria, test reachability, and compute feedback inputs directly from data, including extensions to safe control and regulation. The results enable robust, model-free control of complex logical networks and have potential applications in gene networks and digital systems where data are plentiful but models are hard to identify.
Abstract
In recent years, data-driven approaches have become increasingly pervasive across all areas of control engineering. However, the applications of data-based techniques to Boolean Control Networks (BCNs) are still very limited. In this paper we aim to fill this gap, by exploring the possibility of solving three fundamental control problems, i.e., state feedback stabilization, safe control and output regulation, for a BCN, leveraging only a limited amount of data generated by the network, without knowing or identifying its model.