Title: A coupled reaction-diffusion field model for perception-action cycle with applications to robot navigation

Authors: Eugene Aidman, Vladimir Ivancevic, Andrew Jennings

Addresses: Land Operations Division, Defence Science and Technology Organisation, P.O. Box 1500, Edinburgh SA 5111, Australia. ' Land Operations Division, Defence Science and Technology Organisation, P.O. Box 1500, Edinburgh SA 5111, Australia. ' School of Electrical and Computer Engineering, RMIT University, Melbourne, Australia

Abstract: A generalised reaction-diffusion field model for robot navigation is proposed. It utilises two mutually antagonistic neural fields counteracting in patterns similar to that of flexor/extensor muscles controlling the movements in major joints in the human body. Combining local activation and generalised inhibition represented by Amari|s neural field equations and extended by the Fitzhugh-Nagumo and Wilson-Cowan activator-inhibitor systems, results in the type of neural attractor dynamics that may lead to spontaneous oscillatory pattern formation. Preliminary simulation data suggest that this approach has utility in enabling a team of autonomous vehicles to navigate in a crowded pedestrian crossing.

Keywords: reaction-diffusion systems; coupled activator-inhibitor fields; dynamical grid; robot navigation; crowded environments; local activation; generalised inhibition; neural field equations; neural attractor dynamics; autonomous vehicles; pedestrian crossing; mobile robots; neural networks.

DOI: 10.1504/IJIDSS.2008.021969

International Journal of Intelligent Defence Support Systems, 2008 Vol.1 No.2, pp.93 - 115

Published online: 11 Dec 2008 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article