Premium
Machine‐Learning‐Based Controller Design for Discrete‐Valued Input Systems
Author(s) -
Konaka Eiji
Publication year - 2014
Publication title -
electronics and communications in japan
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.131
H-Index - 13
eISSN - 1942-9541
pISSN - 1942-9533
DOI - 10.1002/ecj.11508
Subject(s) - computer science , control theory (sociology) , support vector machine , control engineering , controller (irrigation) , sequence (biology) , control system , actuator , classifier (uml) , artificial intelligence , control (management) , engineering , genetics , electrical engineering , agronomy , biology
SUMMARY Switching and ON/OFF controls are effective control techniques for control systems equipped with low‐resolution actuators. Such control mechanisms can be modeled as control systems that restrict the control input to discrete values. In this paper, a controller design method based on a machine‐learning technique is discussed. The relationship between the current situation (previous input sequence and previous output sequence), applied input, and output evolution is learned by applying certain machine‐learning methods. Specifically, machine‐learning methods such as the approximate nearest neighbor (ANN) method and support vector machine (SVM) are used in this study. The trained classifier will be a controller that connects the current situation and a suitable control input that can drive the current output to the desired one. The effectiveness of the proposed method is verified for discrete input systems via simulations and experiments.