Characteristic Frequency Input Neural Network for Inertia Identification of Tumbling Space Target

Chuan Ma, Jianping Yuan, Dejia Che

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

Abstract

A novel characteristic frequency input network (CFIN) is investigated for inertia parameters identification of the tumbling space target from the quaternion measurements based on the back propagation neural network. The main innovation of the CFIN is it set the 15-dimensional characteristic frequency vector as the input of the neural network, which is extracted from the constant parameters of the target’s attitude quaternion. The utilization of the characteristic frequency not only reduces the required number of nodes to less than 100, but also improves the learning rate of the neural network. The CFIN is trained using 10000 samples and tested using another 2000 data. It can work well in real-time with very little computational burden and storage space once successfully trained. Moreover, effectiveness analysis illustrates that the CFIN can provide more precise estimation of the inertia parameters than the conventional estimation method, like extend Kalman filter and unscented Kalman filter in most case.

Original languageEnglish
Title of host publicationIntelligent Robotics and Applications - 12th International Conference, ICIRA 2019, Proceedings
EditorsHaibin Yu, Jinguo Liu, Lianqing Liu, Yuwang Liu, Zhaojie Ju, Dalin Zhou
PublisherSpringer Verlag
Pages174-187
Number of pages14
ISBN (Print)9783030275402
DOIs
StatePublished - 2019
Event12th International Conference on Intelligent Robotics and Applications, ICIRA 2019 - Shenyang, China
Duration: 8 Aug 201911 Aug 2019

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11744 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference12th International Conference on Intelligent Robotics and Applications, ICIRA 2019
Country/TerritoryChina
CityShenyang
Period8/08/1911/08/19

Keywords

  • BP neural network
  • Characteristic frequency
  • Inertia identification
  • Tumbling space target

Fingerprint

Dive into the research topics of 'Characteristic Frequency Input Neural Network for Inertia Identification of Tumbling Space Target'. Together they form a unique fingerprint.

Cite this