Multistate network model for the pathfinding problem with a self-recovery property.

Journal: Neural networks : the official journal of the International Neural Network Society
PMID:

Abstract

In this study, we propose a continuous model for a pathfinding system. We consider acyclic graphs whose vertices are connected by unidirectional edges. The proposed model autonomously finds a path connecting two specified vertices, and the path is represented by a stable solution of the proposed model. The system has a self-recovery property, i.e., the system can find a path when one of the connections in the existing path is suddenly terminated. Further, we demonstrate that the appropriate installation of inhibitory interaction improves the search time.

Authors

  • Kei-Ichi Ueda
    Graduate School of Science and Engineering, University of Toyama, Toyama 930-8555, Japan. Electronic address: kueda@sci.u-toyama.ac.jp.
  • Masaaki Yadome
    WPI-AIMR Tohoku University, Sendai 980-8577, Japan.
  • Yasumasa Nishiura
    WPI-AIMR Tohoku University, Sendai 980-8577, Japan.