A new CNN-based method of path planning in dynamic environment

Abstract

In this paper the problem of path planning in a dynamic environment is considered. The minimum cost path is planned using Cellular Neural Network (CNN). CNN is a very useful tool for parallel signal processing and can be implemented using VLSI. In the proposed approach the problem of local minima (dead ends on a map) does not exist. Different criteria can be taken into account during path planning for example: the size of the robot, the traversability cost, the occurrence of dynamic obstacles, etc. The method allows us to specify the goal using semantic labels. The experiments performed in a real static environment and simulations in a dynamic environment have shown the efficiency of the proposed method. © 2012 Springer-Verlag Berlin Heidelberg.