Study on the Learning in Intelligent Control Using Neural Networks Based on Back-Propagation and Differential Evolution

Shenglin Mu, Satoru Shibata, Huimin Lu, Tomonori Yamamoto, Shota Nakashima, Kanya Tanaka

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

In order to obtain good control performance of ultrasonic motors in real applications, a study on the learning in intelligent control using neural networks (NN) based on differential evolution (DE) is reported in this chapter. To overcome the problems of characteristic variation and nonlinearity, an intelligent PID controller combined with DE type NN is studied. In the proposed method, an NN controller is designed for estimating the variation of PID gains, adjusting the control performance in PID controller to minimize the error. The learning of NN is implemented by DE in the update of the NN’s weights. By employing the proposed method, the characteristic changes and nonlinearity of USM can be compensated effectively. The effectiveness of the method is confirmed by experimental results.

Original languageEnglish
Title of host publication4th EAI International Conference on Robotic Sensor Networks
EditorsShenglin Mu, Li Yujie, Huimin Lu
PublisherSpringer Science and Business Media Deutschland GmbH
Pages17-29
Number of pages13
ISBN (Print)9783030704506
DOIs
Publication statusPublished - 2022
Event4th EAI International Conference on Robotic Sensor Networks, 2020 - Virtual, Online
Duration: 21 Nov 202022 Nov 2020

Publication series

NameEAI/Springer Innovations in Communication and Computing
ISSN (Print)2522-8595
ISSN (Electronic)2522-8609

Conference

Conference4th EAI International Conference on Robotic Sensor Networks, 2020
CityVirtual, Online
Period21/11/2022/11/20

Keywords

  • Differential evolution
  • Intelligent control
  • Neural network
  • PID control
  • Ultrasonic motor

Fingerprint

Dive into the research topics of 'Study on the Learning in Intelligent Control Using Neural Networks Based on Back-Propagation and Differential Evolution'. Together they form a unique fingerprint.

Cite this