Evolving Fuzzy Controller, and Application to a Distributed Two-Tank Process

An evolving design approach to fuzzy logic controllers (FLCs) is proposed. The FLC to be evolved is composed of a set of univariate fuzzy control rules, allowing to have a better understanding of the influence of each input variable on the controller's behavior. The criterion to add new control rules is based on a novelty detection criterion allowing to detect data which are considered new, unknown, or badly representative, with respect to the learned FLC. The proposed approach is tested on a Two-Tank process controlled in a networked environment under a proof-of-concept platform, named KhronoSim, for testing cyberphysical systems in closed-loop. The proposed approach has successfully evolved/designed the FLC, controlling the process on unknown (for the controller) regions of operation.