Pilot behavior modeling using LSTM network: A case study

  • Yanan Zhou
  • , Zihao Fu
  • , Guanghong Gong*
  • *Corresponding author for this work

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

Abstract

Traditional behavior modeling methods rely on the knowledge representation derived from the induction and abstraction of subject matter experts, leading to the high barrier and long modeling period. To tackle this problem, we focus on a new behavior modeling approach, which extracts behavior knowledge from behavior data using recurrent neural network (RNN). A case study, take-off behavior modeling using long short-term memory (LSTM) network, was carried out in three phases: the data recording phase, the offline model training phase and the online model execution phase. A three-layer neural network was constructed to learn the pattern of take-off manipulations. The resulting take-off behavior model performed well to ‘pilot’ an airplane in the real-time test.

Original languageEnglish
Title of host publicationTheory, Methodology, Tools and Applications for Modeling and Simulation of Complex Systems - 16th Asia Simulation Conference and SCS Autumn Simulation Multi-Conference, AsiaSim/SCS AutumnSim 2016, Proceedings
EditorsLin Zhang, Xiao Song, Yunjie Wu
PublisherSpringer Verlag
Pages458-465
Number of pages8
ISBN (Print)9789811026652
DOIs
StatePublished - 2016
Event16th Asia Simulation Conference and SCS Autumn Simulation Multi-Conference, AsiaSim/SCS AutumnSim 2016 - Beijing, China
Duration: 8 Oct 201611 Oct 2016

Publication series

NameCommunications in Computer and Information Science
Volume644
ISSN (Print)1865-0929

Conference

Conference16th Asia Simulation Conference and SCS Autumn Simulation Multi-Conference, AsiaSim/SCS AutumnSim 2016
Country/TerritoryChina
CityBeijing
Period8/10/1611/10/16

Keywords

  • Behavior modeling
  • Flight simulation
  • LSTM
  • RNN

Fingerprint

Dive into the research topics of 'Pilot behavior modeling using LSTM network: A case study'. Together they form a unique fingerprint.

Cite this