|
Digital Library of the
European Council for Modelling and Simulation |
Title: |
The Use
Of Artificial Intelligence In Controlling A 6DOF Motion Platform |
Authors: |
Webjørn Rekdalsbakken |
Published in: |
ECMS
2007 Proceedings Edited
by: Ivan Zelinka, Zuzana Oplatkova, Alessandra Orsoni ISBN:
978-0-9553018-2-7 Doi: 10.7148/2007 21st European
Conference on Modelling and Simulation, Prague, June
4-6, 2007 |
Citation
format: |
Rekdalsbakken, W. (2007). The Use Of Artificial
Intelligence In Controlling A 6DOF Motion Platform. ECMS 2007 Proceedings
edited by: I. Zelinka, Z. Oplatkova,
A. Orsoni (pp. 249-254).
European Council for Modeling and Simulation. doi:10.7148/2007-0249. |
DOI: |
http://dx.doi.org/10.7148/2007-0249 |
Abstract: |
Aalesund University College (AUC) has long and broad experience in the design
of nautical simulators. An important part of this work is the design and
control of motion platforms in three or six Degrees of Freedom (DOF). A
small-scale model of a 6DOF motion platform has been designed to investigate
the control possibilities of such a device. The platform deck is moved by six
actuator rods connected to DC servo motors,
controlled by Pulse Width Modulated (PWM) signals. The servo
motors are placed in pairs at the corners of an equilateral triangle
at the base of the platform, and the actuators are similarly connected in
pairs to the platform deck. A solution for the inverse kinematics from
platform coordinates to motor shaft angles is derived in this paper. The
mathematical transformations are verified through simulations and measurements
of the platform movements. To explore the control capabilities of the motion
platform, a ball has been controlled with the aid of a camera to follow
different kinds of paths on the deck. The ball follows reference paths like a
circle or an 8-curve with the platform in a stationary position relative to
the camera, as well as while the platform is moving in a plane at a certain
height. The control is based on a state space model, and different kinds of
adaptive algorithms have been tested, including a Neural Network (NN) and a
Genetic Algorithm (GA). The training of the NN controller with a back
propagation algorithm was very time demanding, and the real-time
implementation was not successful. The GA controller, however, functioned
very well and adapted to the different reference conditions in real-time. |
Full
text: |