Skip to main content
Article
Online Neurocontrol Design Optimized by a Genetic Algorithm for a Multi-trailer System
Transactions of the Society of Instrument and Control Engineers (2006)
  • Endusa Billy Muhando, University of the Ryukyus
  • Hiroshi Kinjo, University of the Ryukyus
  • Eiho Uezato, University of the Ryukyus
  • Tetsuhiko Yamamoto, University of the Ryukyus
Abstract

In this paper we present real-time neurocontrol system for a nonlinear dynamic plant. In order to improve the training and control performance we present a combined system with a neurocontroller (NC) and a linear quadratic regulator (LQR). We apply the control scheme to the backward control of a multi-trailer system. The function of the LQR is to cater for the linear part of the system thereby alleviating the load on the NC. An emulator of the plant is used to design the desired trajectory. The actual plant is subsequently run on this path. In the event that the plant fails to trace the desired trajectory, the control system is re-designed from this point and a new trajectory formulated. We utilize the GA to update the NC weights, while the evaluation function of the NC incorporates both the squared errors and the running steps errors; the latter having the function of realizing faster training of the NC. We have significantly reduced the computation time by utilizing one pattern training for the NCs in real time. Simulations show that the proposed online method has good control performance for the trailer truck system.

Keywords
  • adaptive neurocontrol,
  • multi-trailer system,
  • online design,
  • back-up control,
  • GA training.
Publication Date
September, 2006
Citation Information
Endusa Billy Muhando, Hiroshi Kinjo, Eiho Uezato and Tetsuhiko Yamamoto. "Online Neurocontrol Design Optimized by a Genetic Algorithm for a Multi-trailer System" Transactions of the Society of Instrument and Control Engineers Vol. 42 Iss. 9 (2006)
Available at: http://works.bepress.com/muhando/5/