Home // INFOCOMP 2012, The Second International Conference on Advanced Communications and Computation // View article

A Neural Network Ultrasonic Sensor Simulator for Evolutionary Robotics

Authors:
Christiaan J. Pretorius
Mathys C. du Plessis
Charmain B. Cilliers

Keywords: Robotics; Genetic Algorithms; Neural Networks; Simulators.

Abstract:
Evolutionary Robotics is concerned with using simulated biological evolution to automatically create controllers for robots. Simulation, which reduces the amount of real-world testing, is typically used to accelerate the evolution process. However, the creation of robotic simulators is a difficult and time-consuming process which requires expert knowledge. As an alternative to manual simulator creation, this paper describes the use of Neural Networks to act as simulators for an ultrasonic distance sensor in the Evolutionary Robotics process. The creation of the simulator Neural Networks is discussed and motivated. The simulators are evaluated by means of a comparison with test data. Finally, the simulators are validated by evolving a controller for an obstacle avoiding robot using the simulator Neural Networks. The experimental results show that Neural Networks can indeed be used to simulate an ultrasonic sensor in the Evolutionary Robotics process.

Pages: 54 to 61

Copyright: Copyright (c) IARIA, 2012

Publication date: October 21, 2012

Published in: conference

ISSN: 2308-3484

ISBN: 978-1-61208-226-4

Location: Venice, Italy

Dates: from October 21, 2012 to October 26, 2012